Skip to main content

Full text of "It Wasn.t All Magic: The Early Struggle to Automate Cryptanalysis, 1930s-1960s"

See other formats 

"Rummaging in the government 's attic " 

Description of document: 

Requested date: 
Released date: 
Posted date: 

United States Cryptologic History, Special Series, Volume 
6, It Wasn't All Magic: The Early Struggle to Automate 
Cryptanalysis, 1930s - 1960s , Colin B. Burke, Center For 
Cryptologic History, National Security Agency, 2002 




Source of document: 

National Security Agency 

Attn: FOIA/PA Office (DJ4) 

9800 Savage Road, Suite 6248 

Ft. George G. Meade, MD 20755-6248 

Fax: 443-479-3612 (ATTN: FOIA/PA Office) 

On-line Request Form 

The web site ("the site") is noncommercial and free to the public. The site and materials 
made available on the site, such as this file, are for reference only. The web site and its 
principals have made every effort to make this information as complete and as accurate as possible, however, 
there may be mistakes and omissions, both typographical and in content. The web site and 
its principals shall have neither liability nor responsibility to any person or entity with respect to any loss or 
damage caused, or alleged to have been caused, directly or indirectly, by the information provided on the web site or in this file. The public records published on the site were obtained from 
government agencies using proper legal channels. Each document is identified as to the source. Any concerns 
about the contents of the site should be directed to the agency originating the document in 
question. is not responsible for the contents of documents published on the website. 

Web site design Copyright 2007 




Serial: MDR-68543 
29 May 2013 

In response to your 4 August 2012 declassification request, we have reviewed the NSA 
cryptologic history entitled: It Wasn 't All Magic: The Early Struggle to Automate Cryptanalysis, 
1930s-1960s. The report has been reviewed under the Mandatory Declassification Review (MDR) 
requirements of Executive Order (E.O.) 13526 and is enclosed. Some portions deleted from the 
document were found to be currently and properly classified in accordance with E.O. 13526. The 
information denied meets the criteria for classification as set forth in Section 1 .4 subparagraphs (c) 
and (d) and remains classified TOP SECRET and SECRET per Section 1.2 of E.O. 13526. The 
withheld information is exempt from automatic declassification in accordance with Section 3.3(b) (2) 
and (3) and (h) (6) of the Executive Order. Section 3.5 (c) of E.O. 13526, allows for the protection 
afforded to information under the provisions of law. Therefore, information that would reveal 
NSA/CSS functions and activities has been protected in accordance with Section 6, Public Law 86-36 
(50 U.S. Code 402 note ). 

Since your request for declassification has been denied in part, you may file an appeal to 
the NSA/CSS MDR Appeal Authority. The appeal must be postmarked no later than 60 calendar 
days after the date of the denial letter. The appeal shall be in writing addressed to the NSA/CSS MDR 
Appeal Authority (DJ5), National Security Agency, 9800 Savage Road, STE 6881, Ft George G. 
Meade, MD 20755-6881 . The appeal shall reference the initial denial of access and shall contain, in 
sufficient detail and particularity, the grounds upon which the requester believes the release of 
information is required. The NSA/CSS MDR Appeal Authority will endeavor to respond to the 
appeal within 60 working days after receipt of the appeal. 




Declassification Services 

Encl: a/s 

Cover Photos: (U) (clockwise from upper left): Hypo, Vannevar Bush, Harvest, 
Joseph Desch, Gee Whizzer, Stanford Hooper, Bombe, Samuel 



a a 

a a 

I Special/ SerCe* 1 

i Voiu4fte^6 ja 

a a 

a a 

a @ 

a a 

a e 

a a 

|i 7^^ Early Struggle/ to- A ut&ryi&te/ \ 

a ra 

I Cryptanalytiti 1930y - 1960y \ 

a i 

a a 

a a 

a a 

a a 

[a Colin B. Burke & 

a @ 

i i 

a a 

a a 

a a 

a a 

a a 

a a 

a a 

a a 

a a 


a a 


a a 

I 2002 I 

a a 

a a 

a. a 

TnP cprpPT/frOMIMTffPFI T? WA, i MfT. n i ' i H, nnn '"" '■■■ ""! 

This page intentionally left blank 

Page ii 




(U) Foreword xi 

(U) Introduction 

(U) Before NSA Opened Its Hies 1 

(V) The Stocks Were Not All They i 

(V) Ask What the Agency Did for 2 

(U) Ask Not Whaf the Agency Can Do For 3 

(U) Inspiration and Patience 4 

(V) There Wasn't Enough Magic 5 

(I) A Story with Only a Few Acronyms 5 

(U) Two Decades before the Computer 6 

Chapter 1 (U) An Academic in Need of the Navy ... Until 


(V) An Institution for the Real World 

(U) A Man for All Technologies 8 

(U) More Than an Ingenious Yankee 9 

(U) The Politics of Mathematics and Engineering 9 

(U) The Manager of Science 10 

(V) Bush and Stratton's Dream 10 

(V) Bush Confronts Little Science 10 

(L) Bush's Great Plan 11 

(U) Beyond Analog Mechanical Machines it 

(U) Two Men with a Need 12 

(U) A Man for the Navy 13 

(I') Another Plan for Science and the Navy 14 

(U) Hooper Confronts the Bureaucracy 15 

(U) A Few Men and Women for Secrecy 16 

: (U) The Search for Pure Cryptanalysis 1(1 

(U) From Electronics to Electromechanics 18 

(V) A Young Man for the Future 20 

(U) The Dream Postponed Again 20 

(l) The Dream Reborn, for a Moment 21 

(U) Little Science Meets the Little Navy, Again 21 

(U) A Man for Statistics 22 

(U) Science and the Navy Need Other Friends 23 

(U) The Private World of Science 24 

(U) A Man for Applied Mathematics and Information 24 

(U) American Science and the War - the NDRC 24 


TOP 0CCRi:T//COM I NT//m:L TO uoa, AUO, can, qdr and n zu/x i 

(U) Corporate Charity 25 

(U) The Navy Comes in Second 26 

Chapter 2 (U) The First Electronic Computer: Perhaps 

(U) A Reminder of Hooper's Hopes and Frustrations 33 

(U) The Institutional Context 33 

(U) The First Defeat: Bush Is Rejected 34 

(U) A Machine Too Soon 34 

(U) The Decision to Build a Machine 35 

(U) Bush and Wenger Select a Problem 35 

(U) The Index 36 

(U) An Added Bonus, Possibly 37 

(U) Bush Outlines the Machine and Sets Different Goals 37 

(U) The Comparator Really Doesn't Go to Washington 38 

(U) Too Much to Ask of Mere Machines 39 

(U) No Thanks for the Memories 40 

(U) The Limits of Mechanics 40 

(U) Let There Be Light, But Not Too Much 41 

(U) The Most Difficult Problem of All, But It Wasn't 41 

(U) The Easiest Becomes the Most Difficult 42 

(U) Beyond Murphy's Law 42 

(U) Spring Is a Time for Love, Not Machinery 44 

(U) RAM Project Seems to Die, Late 1938 44 

(U) A Comparator There May Never Be .45 

(U) Big Science Begins to Emerge 45 

(U) Fire Control 46 

(U) The Second Comparator 46 

(U) OP-20-G and Ultra 47 

(U) So Long for So Little 48 

(U) The Search for the Second Comparator 48 

(U) No Equal Partners in Ultra 49 

(U) Another Machine That Wouldn't 49 

(U) The Revenge of Mechanics: the First Rounds 50 

(TS//SI//REL) Logs and Relays - the Gee Whizer 50 

(U) The Navy Get Some Changes 51 

(U) The Greatest Kludge of All, But It Worked 52 

(U) Trying to Save Bush's Reputation 52 

(U) Yet Another Chance 52 

(U) When the Ciphers Can't Be Broken 53 

(U) Wenger to the Rescue 54 

(U) Mathematics to Meet the Great Challenge 54 

(U) Bureaucracy vs. Science, Again 54 

(U) A Seeming Victory for Science 55 



Chapter 3 (U) Bush's Dream Does Not Come True 

(U) A Look Ahead lo Peace 63 

(U) January 1942: Too Much Too Late 63 

(U) A Giant Step Backwards 64 

(U) Haste and Confustion 64 

(U) Tessie Wouldn't FAther 65 

(U) Jessie's New Hat 66 

(U) You Can Use Some of the Technology 

Some of the Time, But 67 

(U) A Machine for Mrs. Driscoll's Special Problem 69 

(U) A Paper War, Perhaps 71 

(U) The Comparator Dies, Again 71 

(U) Almost Another Digital Machine 72 

(U) The Old Technologies Are the Best Technologies, 

for a Time 74 

(U) Meanwhile, the Tabulator's Revenge 74 

(U) LBM's Most Special Contribution 75 

(U) In the Absence of Rapid Machines 77 

Chapter 4 (U) Meeting the Crisis: Ultra and the Bombe 

(U) Looking Ahead - Ultra Saves RAM and OP-20-G 

Creates a Science Company 83 

(U) The "E" Machine 83 

(U) Only a Few Were Able and Willing to Tackle "E" 84 

(U) The Poles Automate Cryptanalysis 

in Their Special Way 85 

(U) Keeping the Bombe Secret for Too Long 85 

(U) A Fresh Start against "E" 86 

(U) Analog and Parallel May Be Fast, But 88 

(U) Ask and Then Not Receive 89 

(U) Gave All and Get 90 

(U) What Happened After 91 

(U) Trust Builds Very Slowly 91 

(U) Agreements and Agreements and Agreements, But 91 

(U) Going Separate Ways 92 

(IT) America without an Ultra 93 

(U) An American Ultra, Perhaps 93 

(U) Faster Than a Speeding Relay 94 

(U) Great British Expectations 94 

(U) Great American Expectations 94 

(U) Trying to Step Forward, Not Back 97 

(U) Britain's Own Version of Bush's Electronic Dreams 97 


top oconny/oo M iNTOncL to uoa, aug, cam, opp. amp hzujxi 

(U) No Time for Electronic 98 

(U)A Crisis of Organizationa and Technology 99 

(U) Searching for a Place in Ultra 99 

(U) The Power of Innocence 99 

(U) The Power of Ignorance 101 

(U) The Cousins Will Have Their Way. a Degree 103 

(U) A Long Apprenticeship 104 

(U) Desch Takes Charge 105 

(U) Wenger Gets His Organization 105 

(U) Of Tires and Transmissions and 

a Disappearing Laboratory 105 

(U) Saving the American Bomhe 108 

(U) A Bombe Too Late 108 

(U) A Program Based on Another Technological Bet 109 

(U) July 26th: a Day of Defeat 109 

(U) A Victory a Bit Too Late 109 

(U) Ignorant No More 111 

(U) The Bombes at Work 111 

(U) More to It Than the Bombe 112 

Chapter 5(U) A Search for Other "Bombes 


(U) Meanwhile, the Army 127 

(U) The Search for Another American Ultra 128 

(U) A Great Electronic Adventure, the Freak 129 

(U) Tabulators and Traffic: A Data Processing War 130 

(U) Making the Tools More Powerful 132 

(U) Slides, Runs, and Endless Decks of Cards 132 

(U) The Other Bombe Program 135 

(U) Another Step Back J.35 

(U) More to It Than the Machine 137 

(U) A Machine Looking for Work 139 

(U) More Emergencies and More Compromises 140 

(U) The Other Purples 142 

(U) New Guys and Old Guys, New Techniques 

and Old Insights 142 

(U) A Matter of Machines and Control 143 

(II) The Snake That Died Too Young, Viper 144 

(U) A Snake in Hand, Perhaps — I*ython 146 

(U) Of Strips and Strippers 148 

(U) Strips Without Strippers 149 

(U) The Attack on the Many JN25S 150 

(U) The Comparators That Weren't the Copperhead 

Proposals and the Victory of Electronics 153 



(I) Beyond the Copperheads - the JN25 Crisis 

and "M's" Response 155 

Chapter 6 (U) Beyond the Bombes and Beyond World War II 

(I) After the Bombe 163 

(U) Every Which Way: The Code Challenge Continues 164 

(U) The Navy's Madame X— the Strongest Selecton 165 

(U) A Wall of Knobs 166 

(V) Walls of Tubes 167 

(U) Into the Beyond and the Past, Rooms of Wires 

and Disk 168 

(U) Desparate Options and a Conservative Selector 168 

(U) Walls of Pipes and Thousands of Dots 169 

(V) The Relay Selector Gets an Electronic Face Lift 171 

(U) The Biggest Snakes of All — The Navy Almost Builds 

an Electronic Bombe 173 

(U) The Serpent and Friends 173 

(V) The Revenge of the Enigma — or Electronics 

Is Inescapable 177 

(U) Beyond Cribs: the Statistical Bombe 178 

(U) No Escaping Electronics, Enigma Meets the Cobra 180 

(U) The Navy's Duenna 180 

(U) From Relays to Tubes, Rosen Gets His Chance 182 

(V) Engineering Pride and Peacetime Priorities 183 

(U) Keeping the Faith: the Return of the Film Machines . . .184 

(V) The Revenge of the Codes, Again 184 

(U) More Numbers Than Ever Before 186 

(U) Dr. Bush, Your Best Friend Is Really the Army 189 

(V) The Great 5202 190 

(U) Beyond the Comparators lgi 

(U) The Machine That Wasn't 191 

Chapter 7(U) The Magic Continues 

(U) Would History Repeal Itself? 199 

(V) What There Wasn't 200 

(U) Signs of Some Appreciation 201 

(V) More MAGIC: Cryptanalysis Continues as Before 202 

(U) A Cryptologic Future: Architecture and Ambiguity 

and Budgets 203 

(U) The Enigma Is Dead (W e Think): 


Long Live the 


(V) A Hangover from Another Time 205 

PL. 86-36 



r.L. ob-ob 

( U) Mrs. O'Malleu' s Wayward So n 206 

(S//SI) The Grand I \Machine 208 

(U) Hecate's Impressive Competitor 211 

(U) The Universal RAMs 21 1 

(U) The Illusive Matrix 212 

(U) It's a Nice Idea, Dr. von Neuman, But 212 

(U) Faith Without Institutions: Slides, Sleds and Skates . . .214 
(U) Faith and an Institution: the Chance to Begin an Era . .218 

(U) A Bright Hope for Hooper's Dream 220 

(U) The Grand Machine of Its Time, the New Comparator . .221 

(U) Meanwhile, a Last Chance for Microfilm 223 

(U) Finally, the Electronic Bombe 223 

Chapter 8 (U) Courage and Chaos: SIGINT and the Computer Revolution 

(V) It Wasn't Safe at the Cutting Edge 233 

(U) An Idea Differed 233 

(U) Goodbye Dr. Bush, Hello Professor von Neuman 236 

(U) A Summer in Philadelphia - an Exciting One 238 

(U) Buy a Computer, Now 241 

(U) Little JTianks for That Memory 244 

(U) Saving a Reputation through Logic 247 

(U) The Army's Problem 248 

(U) Stratum 's Dream Revisited 251 

(U) So Much for Simplicity 253 

(U) Abner's Not Quite Best Friend 254 

(U) Abner by Inertia 255 

(U) Abner's Bad Temper 256 

(U)And Then Came 257 

Chapter 9 (U) Wandering into Ti'ouble 

(U) A Cryptoanalytic Future 263 

(U) The Worst of Times 263 

(U) The Magic Continues 265 

(U) At Last, the Electronic Bombe - Perhaps 266 

(U) Without Magic and without Many Friends 268 

(U) The End of an ERA 269 

(V) ERA's Treasures 270 

(U) SIGINT Loses Another Friend 271 

(U) An Old Friend's Burdens 272 

(U) A Desperate Search for '"Depth" 274 

(U) Wanderers and Nomads and Chaos 276 

Page viii TOP SFrffFT/rfCQM I NTOR EL TO U SA, A U S , CAN , SBR A N D N ZU/X1 


(U) If You Can't Trust Someone from 

the Adams Family, Then 278 

(U) Failure upon Failure 280 

Chapter 10 (U) A Matter of Faith 

(U) Would Science or an Old Tactician Save the Agency? . . .285 

(U) Rushing "Bits," Not Even "-Bytes," into the Agency 285 

(U) Canine Guards the Fort 287 

(U) EnterTom Watson and IBM 289 

(U) A Machine for Us, Perhaps 292 

(U) One Big Machine Beats Out Many Little Ones 293 

(U) An ERA by Any Other Name Is IBM 295 

(U) Is Half a Farmer Better Than 296 

(U) St. Paul in Mohansic 297 

(U) Bucks Talk: the Favored Sister Gets the Attention 297 

(U) A Data Factory 298 

(U) Not a Farmer, a Nomad 299 

(U) Engineering Is Not Science, at least to the SAB 300 

(U) You Can Take Science Out of the Agency, But Can You .302 
(U) Almost in Science - Would Lightning Be 

the Other Harvest 305 

(U) What Kind of Friend Are You, Dr. Baker? 307 

(U) Dr. Baker's Half-an-Institute 309 

(U)A Harvest ofOverexpectations 310 

(U) ERA's and the Shop -floor Cryppies 7 Revenge 313 

(U) Technology and Faith, 1962 316 



This page intentionally left blank 




mm mmm mmm - •• •* »■ • *• * >~ g&mm$tm 

Y.^.A :- 3 '»■,.# WiWxusp** 

(U) Conventional wisdom about NSA and com- 
puters has it, as a retired NSA senior officer once 
wrote me, "In the early days, NSA and its predeces- 
sor organizations drove the computer industry. In 
the 1960s, we kept pace with it. We started losing 
ground in the '70s, and in the '80s we struggled to 
keep up with the industry." 

(U) True, but underlying this, in each decade 
the cryptologic organizations experienced a wide 
range of successes and failures, positives and nega- 
tives. If, as slang puts it, "they won some, lost some, 
and some got rained out," all of this experience is 
worth serious examination by students of comput- 
ers, cryptanalysis, and NSA history. 

(U) The current volume, Dr. Colin Burke's It 
Wasn't All Magic: The Early Struggle to Automate 
Cryptanalysis, 193OS-1960S, contains a view of the 
first decades of computer development that is 
broad and deep and rich. 

(U) It begins in the 1930s as American and 
British intelligence officials confronted new crypt- 
analytic and cryptographic challenges, and adapted 
some intriguing new concepts to their analysis. It 
carries the story to the flexible and fast systems of 
the late 1950s and early 1960s. 

(U) Dr. Burke follows and links the develop- 
ment of automatic data processing from the critical 
conceptual work of the 1930s through the practical 
experiments born of national necessity in the world 
war to the postwar development and the previously 
untold story of NSA's postwar computer develop- 
ment. Along the way, he has rescued from obscuri- 
ty some important successes - and some important 
failures - in cryptanalytic machinery from World 
War II. 

(U) All too often, discussions of NSAs comput- 
er development treat only the mainstream, ignoring 
the problems, failures, dead ends and might-have- 
beens, in order to concentrate on successes. In the 
present volume, however, key components of 
Dr. Burke's story and important for our knowledge 
are the machines which didn't work or which never 
had progeny, and why this was so. Just as impor- 
tant are Dr. Burke's cautionary tales about the 
influence of international and interservice rivalry 
on plans and procedures. Technical limitations and 
technical opportunities shaped much of the devel- 
opment of computing equipment, but the story is 
also replete with instances of man-made barriers 
and baleful bureaucratic bypaths that wielded great 
influence during much of this development. 

(U) A word about how this manuscript came to 


(U) A no less important factor than the infor- 
mation and analysis in this current volume is that it 
represents an objective view of NSAs computer his- 
tory by a writer not from NSA or one of the Service 
Cryptologic Elements. The author, a university pro- 
fessor, had no stake in either defending or 
besmirching decisions made fifty years ago or the 
organizations or people who made them. 

(U) The Center for Cryptologic History, 
between 1990 and 1999, administered a program to 
bring outside academics or researchers to the 
CCH for special projects. The CCH sponsored six 
Scholars in Residence in that period; of the six, the 
first two received security clearances for work on 
classified projects. The others remained uncleared 
and worked only with declassified materials. 

(U/ /FQUO) Dr. Colin B. Burke was the second 
of the two cleared scholars. 


Page xi 


(UZ/ FOUO^ Dr. Burke's professional biography 
is given at the end of the book. Note that as a pro- 
fessor of both history and computer research tech- 
niques, Dr. Burke became one of the pioneers in the 
field of computer history. 

(U/ /fOUQj As background reading, T recom- 
mend not only Dr. Burke's own unclassified 
publications, available commercially, but two clas- 
sified histories available from the Center for 
Cryptologic History: Thomas R. Johnson, 
American Cryptology during the Cold War (4 vols) 
and Michael L. Peterson, BOURBON to Black 



Center for Cryptologic History 



( U) Introduction 

(U) Before NSA Opened Its Files 

(U) I am one of those "outsiders" I talk about 
so much in the later chapters of this book. I was 
fortunate to be brought into the National Security 
Agency as one of the Center for Cryptologic 
History's first Scholars in Residence. I was bor- 
rowed from my university because I had spent a 
decade working on the history of computers at 
NSA's predecessors. I even had the courage to 
write a book about the subject. 1 

(U) That monograph was on the machines, 
policies, and relationships that led to the U.S. 
Navy's cryptanalytic machine (computer) pro- 
gram in World War II. The book was also about 
the first major attempts to automate the 
American library. It had to be about both because 
the same people built bibliographic and cryptan- 
alytic machines. 

(U) My study covered events in the history of 
the machines through the 1940s, but its focus was 
on the period between 1930 and 1945. An impor- 
tant conclusion was that the relationship between 
the efforts of America's codebreakers and the 
emergence of the modern digital electronic com- 
puter was more complex than had been thought. 
The navy's cryptanalysts were in a push-me, pull- 
me situation. Their work made mechanization a 
necessity, but the pressures of war and the refusal 
of the government bureaucracy to sponsor long- 
term research and development programs pre- 
vented the navy from becoming the inventor of 
the modern computer. 

DuUid Piuni. NOA/COSM ih;i m 

Dated 2.4 I\-l>i uai y 199O 
DulaAsify On. Xi — 

(U) During World War II American cryptana- 
lysts built some of the most sophisticated elec- 
tronic machines in the world, but the need to 
address cryptanalytic crises blocked them from 
creating the general-purpose digital electronic 

(U) Just as my book was published, I was 
asked to come to the National Security Agency. 
One purpose of my year in residence was to see if 
it was possible to write a complete history of com- 
puters at the Agency. The goal was a monograph 
that covered the entire life of NSA and its prede- 
cessors. The thought of finally being able to see 
the many highly classified documents that had 
been withheld from me more than balanced the 
pledge I had to give: I had to promise to refrain 
from publishing without the approval of NSA's 

(U) The Stacks Were Not All They... 

(U) I began my residency by surveying the 
Agency's archive holdings and by rereading the 
few synthetic works that had been declassified. 
The comprehensiveness of the archive holdings 
was critical because unless enough of the correct 
type of documents had been saved and indexed, 
there was little chance to produce a history of the 
post- World War II era. Useful documents from 
the 1950s through the 1980s were of special 
importance because the Agency had allowed 
almost nothing about its operations in the last 
forty years to be made public. There was not even 
a counterpart to the informative but very 
unhandy collection of documents on the pre-1946 
period released to the National Archives, the 
Special Research History series. 

(U) As I examined the collections, I was 
pleased to find that the materials I needed to 


Page 1 


revise my earlier work were plentiful and well 
organized. There were several contemporary 
studies of computer efforts up to 1945 that were 
technically as well as historically enlightening, 

(U) I was also happy to discover that although 
I had not been correct in all the details about the 
early American cryptanalytic machines, I had 
come quite close. I concluded that I had drawn an 
acceptable overview of the computer efforts of the 
army and navy cryptologic agencies up to 1950. 

(U) But I was less than content with the mate- 
rials for the postwar era. I decided that it would 
be impossible to tell the type of story about the 
forty-five years since the Korean War that I had 
done for the earlier period, at least not within one 

(U) However, there were many documents, 
several oral histories, and the work of Samuel S. 
Snyder to provide a basis for a history of comput- 
ers within NSA up to the early 1960s. 

(U) Ask What the Agency Did for... 

(U) Samuel Snyder was one of the founding 
fathers of machine cryptanalysis. He joined the 
army's cryptologic unit in the 1930s and 
remained at the Agency, becoming important to 
many of its computer projects of the 1940s and 
1950s. His experience, his desire to document the 
Agency's computer history, and his commitment 
to the welfare of NSA made its administrators 
receptive to his requests to be allowed the time 
and resources during the late 1960s and early 
1970s to write about the Agency's computer his- 

(U) Mr. Snyder completed several works. 
They ranged from sketches of the history of the 
army's first electronic computer, Abner, to a sur- 
vey of the general-purpose electronic computers 
the Agency had built or purchased. The study of 
the K gp" computers seemed so important and fit 
so well with the Agency's desire to obtain good 

publicity that it was declassified and, in various 
forms, published in the open literature. 2 

(U) Mr. Snyder's work showed the contribu- 
tions of NSA to the development of computer 
technology and to the emergence of the American 
computer industry. It made it clear that NSA had 
been a major sponsor of technical advances. like 
several other large government agencies, its com- 
puter purchases and its research and develop- 
ment contracts helped establish America as the 
world's leading computer manufacturer. He also 
made it clear that NSA had been at the cutting 
edge of computer technology and architecture. 

(U) I was very tempted to just deepen Sam 
Snyder's work on post-1950 automation, but as I 
went through the hundreds of record boxes at the 
archives and as T began to reflect on their con- 
tents, I decided that I had to do something differ- 
ent. I had to take an alternative view of the 
Agency's computer history. 

(U) Because of the need to guard NSA's crypt- 
analytic methods, Mr. Snyder could not discuss 
the reasons for the development of the devices he 
included in his published works. Just as the 
Agency cannot reveal its cryptologic successes 
without endangering them, Snyder could not give 
either the "why" or, in many cases, the "what" of 
the Agency computers to the public. The jobs the 
NSA computers had to perform and the decision 
processes that led them to be part of the famous 
collection of machines that once resided in the 
Agency's "basement" had to be left out of his stud- 

(U) He certainly could not discuss what were 
and are the most intriguing machines at NSA, its 
dozens of special-purpose computers, ones whose 
architecture embodies a cryptanalytic process. 
Doing more than listing their cover names would 
have revealed what methods the Agency was 
using and what targets it was attacking. Neither 
the NSA nor the British intelligence agencies were 
responsible for the initial release of information 

Page 2 



about MAGIC and ULTRA. The stories were told 
by others. 

(U) The restrictions on Mr. Snyder had anoth- 
er influence. Because he was forced to divorce 
SIGINT and computer history, the machines he 
described seemed to have emerged as a result of 
an indestructible synergy between the Agency, 
the computer industry, and America's scientific 

(U) My knowledge of the course of technolog- 
ical advancement in the cryptologic arena in the 
1930s and 1940s led me to doubt that the postwar 
era's machine history was so smooth and problem 

(U) There was something more fundamental 
about Mr. Snyder's approach that led me to 
search for alternative ways to interpret the docu- 
ments that were emerging from the NSA archives 
and the offices of old hands at the Agency. 
Although Snyder's articles are invaluable, his 
interpretation seemed unlikely to be able to bring 
together the policy, the cryptanalytic, and techno- 
logical histories of SIGINT computerization. 

(U) The emphasis in his public articles was on 
what the Agency's computer efforts did for others, 
especially the computer industry. But I had a 
clear sense that I could tell the story of computers 
only by using an approach that was the near 
opposite of Sam Snyder's. Focusing onNSA's role 
in transferring technology and supporting com- 
mercial computer development, I concluded, hid 
as much or more than it revealed. 

(U) Ask Not What the Agency Can Do for... 

(U) To understand computerization at NSA 
the question should not be "what did NSA do for 
the computer industry?" but "what was it that the 
industry could not or would not do for NSA?" 

(U) The second question leads an investigator 
into the many technical, institutional, and politi- 

cal struggles that the Agency faced as it attempt- 
ed to keep up with the cryptologic capabilities of 
its adversaries. It helps to explain why NSA has 
had to build so many special computers, why it 
invested in several technological misadventures, 
and why it had to involve itself in some unusual 
relationships with private industry and academia. 
The question also helps to integrate the political 
history of NSA with its drive to advance mathe- 
matical cryptanalysis. 

(U//FOUO) For example, using the "couldn't 
do" approach to Agency projects led me to an 
understanding ofNSA's great computer adven- 
ture, the Harvest system. It also helped uncover 
the Agency's reason for creating its high-powered 
mathematical think tank at Princeton. Both the 
computer and the institution were the result of 
much more than a desire to extend the reach of 
formal cryptanalysis; they were born of intense 
political pressures on the Agency, and they were 
grand compromises rather than perfect solutions 
to abstract problems. 

(U) Spotlighting what the computer industry 
could not do for signals intelligence also helps to 
integrate the Agency's computer and cryptanalyt- 
ic histories. Putting them together shows that 
inventing and developing an effective technology 
for SIGINT has been difficult and, at times, ago- 

( T0//0I//REL) The "couldn't do" question 
also illuminates one of the most fascinating 
aspects of the history of computers at the 
National Security Agency: the drive to define and 
implement a computer architecture that was rad- 
ically different from the classic design that is now 
called the von Neumann architecture. Since at 
least 1946, American cryptanalysts have done 
much more than use parallel and pipeline pro- 
cessing; they have sought, and came close to 
achieving, a unique architecture for a cryptanalyt- 
ic general-purpose computer. Because the com- 
puter industry was unwilling to develop machines 
that served only one customer, NSA was forced to 


Page 3 


design and build not only single-purpose 
machines but its own versions of a general-pur- 
pose computer. Its Sled and Dervish family of 
high-speed devices reflects the Agency's special 
needs and challenges. Unfortunately, creating an 
all-purpose special NSA computer was too much 
for the Agency, given its resources. Its famed 
Harvest computer, for example, was only a partial 
representation of a tme "cryptanalytic" computer, 
partially because the computer industry could not 
focus on Agency needs. 

(U) When the "couldn't do" question is 
extended to American universities, it becomes 
easier to understand the difficult relationship 
between NSA and "outside" scientists and their 
institutions. Finding and utilizing academic tal- 
ent was very difficult for the Agency. Professors 
did not rush to NSA before or after World War II, 
and they did not pursue much research that was 
of direct help to operational cryptanalysis. 
Devising ways to preserve the Agency's secrets 
and its independence while channeling the con- 
tributions of academics proved to be very diffi- 

(U) Inspiration and Patience 

(U) Something besides the relationships with 
the commercial computer industry and academia 
has to be called on to explain NSA's computer his- 
tory. To show why the Agency created its own 
computer architecture and why it accepted some 
technological retrogressions calls for a bit of 
cryptanalytic muckraking. The craft of code and 
cipher breaking has to be stripped of its romanti- 

(U) The stories of the successful American 
attack on Japan's diplomatic ciphers before 
World War II (PURPLE-MAGIC) and the tri- 
umph over the German ENIGMA-ULTRA are 
used as popular models for the way cryptanalytic 
work proceeds. The popular view shares much, in 
terms of fundamentals, with the public view of 
Agency computer development. 

(U) In the typical story, codebreaking success- 
es such as MAGIC and ULTRA came about 
because of the quick and intense work of a hand- 
ful of geniuses armed with brain power and little 
else. The common image is that once the likes of 
William F. Friedman and Alan Turing had their 
flashes of insight, a flow of precious information 
(and only valuable information) was captured, 
processed by a hastily constructed but ingenious 
machine, and then directed to decision makers. 

(U) Such an image of a heroic cryptanalysis is 
far from being true or useful. Cryptanalytic and 
technological victories have not come as easily as 
that. Even during the glorious codebreaking days 
of World War II, America's cryptanalysts barely 
kept up with their enemies. 

(U) There have been moments when great 
breakthroughs have led to critically important 
messages. And the penetration of some systems, 
such as the U-boat "E," led to a stream of imme- 
diately important information. But typical crypt- 
analysis was and remains a continuing struggle to 
discover patterns and to make sense out of moun- 
tains of raw data. 

(U) Most cryptanalytic solutions have come 
only after years of the most tedious and disdain- 
ful work. The intellectuals that Britain gathered at 
Bletchley Park had to perform mind-deadening 
menial computing tasks hoping that all their 
labor would reveal mistakes by the German cryp- 
tographers and patterns within message texts. 
They had no magical mathematical formula that 
eliminated the need for massive data processing. 
Even when a system was penetrated, creating 
useful information from intercepts called for 
large-scale data handling. 

(8//3I) During the first decades of the Cold 
War, when America's enemies made the ENIG- 
MA systems seem like cryptanalytic child's play, 
NSA could not re-create a MAGIC. It had to wring 
information out of traffic analysis, plain text, and 
even clear voice messages. That forced the Agency 

Page 4 


t o p aeeneTOeoM i NTO R iiL to uoa, auo, oa n , odr a n d n zl//xi 

to become one of the world's largest data proces- 

(U) On top of the special computing needs of 
cryptanalysis, NSA's insatiable need for what 
many times was unique data processing equip- 
ment, made it a computer leader. 

(U) The National Security Agency has not 
been such an important influence in computer 
development because of its mathematical wiz- 
ardry or because it has a mandate to transfer 
technology to the private sector. The Agency's 
contributions have come because of the unique 
nature of cryptanalysis and SIGINT and the 
increasing difficulty of fulfilling a central respon- 
sibility: the production of signals intelligence. 
The Agency has sponsored supercomputers for 
mathematics since the 1940s, but so have many 
others. The critical contributions of the Agency 
have come because of the special needs of opera- 
tional cryptanalysis and SIGINT data processing. 

(U) There Wasn't Enough Magic 

(U) In addition to the "couldn't do" perspec- 
tive, a useful way to understand the history of 
NSA's computers is to place them within the con- 
text of the struggles to overcome the particular 
machines and methods of America's determined 
and increasingly clever opponents. 

(U//FOUQ) The National Security Agency 
has never bought or built computers for abstract 
reasons. Its computers, even those for its hush- 
hush think tank at Princeton, were acquired to 
respond to very practical and immediate needs 
and opportunities. From the early 1930s, when 
the first IBM tabulating machines were brought 
into the secret rooms of the Navy's OP-20-G, to 
NSA's massive computer projects of the 1950s, 
and to the 1990s when the NSA computer build- 
ing is filled with massively parallel and pipe-lined 
special-purpose computers, the Agency's 
machines have been for the solution of problems. 
There have been moments when the Agency has 

been allowed a bit of a luxury to pursue long-term 
and general technological explorations, but they 
have been rare and were always under the threat 
posed by a shift in national, political, or military 

(U) Despite all that, NSA has arguably been 
the largest single user of advanced computing 
machines in the world. It had to be. And because 
of the unique problems it faced and methods it 
used, it also became one of the most sophisticat- 
ed sponsors of new computer and electronic 
equipment. To do its job it had to invest hundreds 
of millions of dollars into research and develop- 

(U) A Story with Only a Few Acronyms 

(U) The story of the Agency's struggle for 
automation from 1930 to the beginning of the 
1960s could become an exercise in the use of 
acronyms. NSA and its predecessors were 
bureaucracies with dozens of subdivisions and 
name changes. The designation for the army's 
crypto branch, for example, was altered several 
times before the end of World War II. Following 
such changes is too much to ask of a reader who 
wishes to gain the "big picture" of cryptanalytic 
computer history. 

(U) To keep the text readable, I decided to use 
as few names as possible for agencies and their 
subdivisions. For example, the army's cryptana- 
lytic branch is called the SIS until the formation 
of NSA 

(U) I have also kept the goal of readability in 
mind when describing machines and processes. I 
have tried to use common terms whenever possi- 
ble, even at the cost of glossing over some techni- 
cal distinctions. I have even used the terms SIG- 
INT and COMINT interchangeably, except in the 
contexts in which the differences between the two 
are significant to understanding NSA's history. 

TOP S E CR E TOe O M I N T /m E L TO U3A, AU3, CA N , ODR A N D N ZU/X 1 

Page 5 


(V) Two Decades before the Computer 

(U) Notes 

(U) I have also tried to use consistent termi- 
nology, although the story of the struggle to auto- 
mate American cryptanalysis begins two decades 
before the modem electronic digital computer 
emerged, NSA's automation story begins in 1930 
when a bright and devoted navy man, Stanford 
Caldwell Hooper, realized that mechanized 
cipher-making was outpacing cryptanalysis. In 
his attempt to modernize the navy's cryptanalytic 
branch, OP-20-G, he engaged a problem that 
proved difficult for America's codebreakers for 
over a generation: How can a secret agency find 
and use the best talents and technology in the 
outside world? Hooper tried to solve that prob- 
lem by creating anew type of relationship with 
academics, specifically the man who became the 
czar of America's new Big Science of World War 
II and the first years of the Cold War - Vannevar 
Bush of MIT. 

1. (U) Colin B. Burke, Information and Secrecy: 
Vannevar Bush, Ultra, and the Other Memex, 
Metuchen, NJ, & London: The Scarecrow Press, 1994. 

2. (U) For example, Samuel S. Snyder, "Computer 
Advances Pioneered by Cryptologic Organizations," 
Annals of the History of Computing , 2 (1980): 61. 

Colin Burke 
December 1994 

Page 6 

top 3 i :cri:t/<oom i nt//r i :l to uoa, aus, can, odr and nzu/x i 


Chapter l 
(U) An Academic in Need of the Navy ... Until 

(U) America's communications intelligence 
services were even more dependent on outsiders 
during the 1930s than they were during the Cold 
War. Neither the army nor the navy had enough 
resources to be technological innovators. They 
could not afford their own research and develop- 
ment programs, and they did not have enough 
money to tempt the scientific and industrial sec- 
tor into concentrating on the cause of advanced 
military technology. All the army and navy 
branches suffered, but those whose functions 
were not highly valued by the military found 
innovation far beyond their reach. 
Communications intelligence was among the dis- 

(U) When it appeared that technology was 
about to outrun the established cryptanalytic 
methods, the American army and navy's commu- 
nications intelligence services had to find ways to 
compensate for their inadequate budgets and the 
absence of relevant research and development 
departments within the military. 

(U) The navy was the first to try to muster new 
technologies to conquer advances in code and 
cipher systems. As early as 1930, one of its more 
progressive leaders, Stanford Hooper, sought 
ways to overcome the financial and organization- 
al constraints on innovation in cryptanalysis. The 
barriers were so great that Hooper could not take 
a direct route to the creation or even the acquisi- 
tion of modern calculating and data processing 
instruments. He did not have the funds to under- 
write an independent development project within 
the navy or within the leading corporations. He 
had to do the best he could with what help could 
be obtained and do it without obligating the navy 
to any major financial or institutional burdens. 

(U) Hooper's odyssey led him to all those he 
thought might be willing to give the navy help 
without asking too much in return. Among the 
many contacts he made during the 1930s, one 
was of special importance to the history of com- 
puters and the cryptanalytic rapid analytical 
machines (RAM). Hooper was able to make an 
arrangement with Vannevar Bush of the 
Massachusetts Institute of Technology. 

(U) To understand the complex, near byzan- 
tine histories of the development of computers 
for cryptanalysis, especially the path-breaking 
RAM program, and to appreciate the difficulties 
of linking science and codebreaking, the biogra- 
phy of one of the leading high-tech universities is 
required. As well, the life of one of the most 
important figures in the history of applied sci- 
ence, Vannevar Bush, needs to be sketched. Of 
special importance is the work Bush had begun 
for other purposes. His attempts to build innova- 
tive machines for scientific calculation and for 
data retrieval determined what technologies he 
recommended as the basis for the first modern 
cryptanalytic machines. 

(U) An Institution for the Real World 

(U) Since its birth on the eve of the Civil War, 
the Massachusetts Institute of Technology had 
been devoted to applying science to practical 
affairs. Its founders rejected much of the curricu- 
la of the traditional American liberal arts college 
as well as the simple vocational program of the 
trade school. They made MIT an example of how 
practical men who worked in cooperation with 
the new institutions of science and industry could 
turn a university into a force for positive change. 
Its founders, such as William Barton Rogers, 
wanted to create men of vision, men who would 
bring the benefits of technology to a backward 


Page 7 


America. Rogers and his colleagues were the 
buDders of some of the most important instru- 
ments of the American industrial revolution. 
Their famous Comparator, for example, allowed 
the exact replication of mass-produced parts. 1 
The name "Comparator" was probably selected 
for Bush's 1930s cryptanalytic machine because 
of the earlier MIT device created by the Institute's 

(U) Rogers initiated MITs long-term policy of 
conducting research for business and govern- 
ment agencies and of having its faculty actively 
engaged in technological and business efforts. 
MIT was badly hit by the recession of the 1870s, 
and its future remained unsure until the water- 
shed years of American life in the 1890s. Then, 
with a more secure financial condition and a 
growing body of alumni and friends who had ben- 
efited from the work of its faculty, the Institute 
expanded its curricula, acquired modern equip- 
ment, and established itself as a force in 
American academic and industrial life. The 
Institute gained a solid reputation in civil and 
mechanical engineering, architecture, naval con- 
struction, chemistry, and electrical engineering. 
By the 1920s the sparkling electrical engineering 
department added a focus on the new fields of 
electronics and communications. The inaugura- 
tion of Samuel W. Stratton as MITs president in 
1923 accelerated the shift to electronics and rein- 
vigorated the school's attempts to create meas- 
urement devices for industry and science. 
Stratton's background and interests blended with 
those of MIT. Stratton's interests and goals fit 
with those of the Institute's faculty, especially 
some of the younger men who sought administra- 
tive approval of their visions for MIT. All at the 
Institute seemed to agree that more support 
should be given to research, and most hoped that 
the school would become a center for the applica- 
tion of formal mathematics to engineering prob- 
lems. One of the junior faculty with such a hope 
was Vannevar Bush. The harmony between 
Bush's and Stratton's views had much to do with 
the younger man's success. Bush received critical 

support from Stratton, allowing him to become 
one of the most important men in the history of 
American science and technology. 

(U) A Man for All Technologies 

(U) A generation later, at the end of World 
War II, Vannevar Bush 
was one of the most 
powerful scien- 
1^ tists the world 
had ever 

known and a 
man familiar 
to most 
The heritage 
of his poli- 
cies contin- 
ues to shape 
\WJ the organiza- 
'* f ^^*j-:*~ tion of academic 
research in 

America. Although 
M vannevar his plan for a federal 

role in science was not completely fulfilled, the 
National Science Foundation is testimony to his 
influence. 2 Bush was important because of his 
influence in such matters as the beginnings of the 
atomic bomb project and the establishment of the 
National Defense Research Committee (NDRC) 
and the National Science Foundation. Despite his 
enormous contributions while at MIT, despite his 
influence within the inner circles during World 
War II and the Cold War, and despite his role in 
shaping the nature of Big Science, and thus the 
modern American university, little was written 
about him until very recently. 3 

(U) The new interest has taken a rather unex- 
pected turn. Instead of focusing on his policy con- 
tributions, the spotlight has been on Bush's role 
in the emergence of computers and information 
processing. The research on his contributions to 
computers arose as the new field of computer his- 
tory was born in the 1980s. The seemingly more 

Page 8 



intense interest in his role in the birth of informa- 
tion science was generated by the rediscovery of 
Bush's work on automatic data retrieval. His 
ideas for a mind machine, Memex, are now treat- 
ed as the origin of hypertext and similar knowl- 
edge systems/ 

(U) More Than an Ingenious Yankee 

(U) Bush merged science with tinkering, if not 
technology. He was an inventor and a natural at 
putting technology together in different combina- 
tions to fulfill a need. His efforts were always 
goal-oriented because he realized that inventions 
required a market to be successful. Bush paid 
attention to the commercial aspects of technology 
and built an enviable list of patents on devices 
ranging from thermostats and typewriters to elec- 
tronics. 3 

(U) After receiving bachelor's and master's 
degrees in engineering from Tufts, Bush gained 
some shop-floor experience while working for the 
giant General Electric Corporation. He pushed 
himself to complete a joint Harvard-MIT doctor- 
al program in electrical engineering. He suffered 
through a great deal of tedious calculation for his 
thesis. Naturally, he searched for shortcuts to 
complete his mathematical analysis of complex 
electrical circuits and applied some of the many 
tricks mathematicians used before the advent of 
the modern computer. The doctoral degree and 
the favorable impression Bush made on Dugald 
Jackson of MIT soon proved of great value. 6 

(U) Just out of school, Bush became associat- 
ed with World War I's New London Research 
Laboratory where the famous Robert Millikan 
brought the nation's best men to focus science on 
the critical U-boat threat. Bush contributed to the 
research with a significant detection system, 
joined the Naval Reserve, and became at least a 
junior member of the national military-scientific 

(U) A complex path eventually led to Bush's 
being one of the creators of Raytheon, a company 
that was able to challenge RCA's patent monopoly 
over radio. Raytheon became one of the many 
important companies tied to MIT and its stu- 
dents. 7 Bush's postwar entrepreneurial ventures 
did not end his academic ambitions, however. He 
accepted a position as an assistant professor in 
MITs Electrical Engineering Department. It was 
understood that he would concentrate on the 
problems of high-power transmission, a focus 
that was sure to attract support from the private 
power companies which were beginning to con- 
struct large regional networks. A string of articles 
on power problems and the mathematical tech- 
niques useful for their solution advanced his aca- 
demic standing. 

(U) The Politics of Mathematics and 

(U) Although Bush was a practical man, he 
was also a missionary for the application of math- 
ematics to engineering and science. He realized 
he had limited formal mathematical skills, but he 
compensated by supporting the work of men like 
the renowned Norbert Wiener. Wiener was 
brought to MIT to integrate advanced mathemat- 
ics with teaching and engineering research. Bush 
also encouraged his students to expand the fron- 
tiers of mathematical engineering, with some 
great results. Claude Shannon, a father of mathe- 
matical information theory, was one of the many 
young men influenced by Bush and Wiener. Bush 
successfully courted the leaders of almost every 
high-tech related corporation in America. 
General Electric, Eastman, NCR, General Motors 
and many other large corporations were familiar 
Bush stomping grounds. Significant yet unex- 
plained, Bush did not develop cordial and prof- 
itable connections with the two major manufac- 
turers of calculating equipment, IBM and 
Remington Rand. History would have been dif- 
ferent if IBM had chosen MIT over Columbia and 
Harvard Universities for its attention and if 


Page 9 


Remington's leader had made a commitment to 
academic research. 

(U) Within a decade after his MIT appoint- 
ment. Bush was a member of the most important 
scientific organizations. Although the United 
States did not have truly powerful scientific insti- 
tutions, ones with the financial resources to shape 
the course of research, such bodies as the 
National Academy of Sciences, the National 
Research Council, and the National Advisory 
Committee for Aeronautics could influence what 
science policy there was. They also provided 
invaluable contacts for their members. By his late 
forties, Bush had become more than a member of 
such groups. He was a statesman of American sci- 

(U) The Manager of Science 

(U) After a long stint as dean and then as vice 
president of MIT, Bush became a significant 
national influence. In 1938 he became the head of 
the Carnegie Institution, one of the most impor- 
tant scientific research agencies in the world. 
That led to his assuming the. leadership of World 
War II's very powerful National Defense 
Research Committee. The NDRC filtered hun- 
dreds of millions of dollars of government funds 
to privately directed research for the war effort. 8 
The NDRC was an improved and vastly expanded 
version of the World War I submarine project and 
was the fulfillment of some of Bush's long-held 
dreams about research in America. The NDRC 
allowed academics something close to the best of 
all worlds: They received government funding 
free of most bureaucratic direction. It also fit with 
Bush's belief that the military would change only 
in response to outside pressures. By the end of the 
war, Bush was the most powerful man in 
American science and was a force the military had 
to recognize. 9 

(U) Bush and Stratton's Dream 

(U) Bush began his work at MIT with research 
on electrical systems. In the early 1920s, Bush 
directed his students to expand the reach of ana- 
log computing. They began with rather simple 
combinations of rods and gears to create 
machines for the automatic calculation of differ- 
ential equations, but those first integraphs were 
more than extensions of the old wire and cone 
contraption that had made Stratton's reputation. 
The young men edged towards solving the major 
mechanical problems that had prevented the 
engineer's friend, the planimeter, from becoming 
a truly powerful tool. By the late 1920s, Bush and 
his men were convinced they had overcome the 
critical problem of torque. They persuaded 
Stratton and the other influentials at MIT that a 
new and startling version of Lord Kelvin's 
machines could be constructed and put to pro- 
ductive use in a few years. 

(U) Bush was allowed to assign the best grad- 
uate students to the creation of the Differential 
Analyser. In 1931, he announced to the scientific 
community that the world's largest and most 
powerful calculating machine stood ready at MIT 
to advance science and engineering. It brought 
international fame to Bush and MIT. 
International visitors came to the Institute and 
clones were built in Europe and America. 
Aberdeen Proving Grounds and the University of 
Pennsylvania built versions, and General Electric 
found it so useful it invested in a copy for itself. 

(U) Bush Confronts Little Science 

(U) Just as Bush's Analyser was given so 
much by the Institute in the late 1920s, the school 
lost its state subsidy. Worse, Stratton's hopes that 
America's largest corporations would donate a 
constant stream of funds to MIT proved unrealis- 
tic. MIT found it more and more difficult to 
finance research with its own resources, and its 
leaders feared that it might be. forced to retreat to 
the vocational model of technical education. The 

Page 10 



faculty, including Vannevar Bush, was on its own 
and all had to struggle for the means to continue 
research and to finance their graduate students. 
The Institute's new president of 1930, Karl T. 
Compton, was as much apart of elite science as 
Stratton, but he was more academic in orienta- 
tion. A famous physicist, Compton arrived with a 
mandate to turn the Institute back towards a true 
scientific curricula and to integrate the latest sci- 
ence with both teaching and research. Wishing to 
reduce the growing ethical and educational prob- 
lems stemming from the staffs business activi- 
ties, and hoping to secure the funds needed to 
allow internal financing of research, Compton let 
it be known that he desired more effort for the 
Institute and less for faculty pocketbooks and cor- 
porate sponsors. 10 Informally, faculty were asked 
to conduct research of general, not particular, 
import. Formally, consulting fees were to be 
shared with the Institute, and patents were to 
become the property of the school if the work had 
been internally funded. To control the increasing- 
ly complex patent problems and to avoid the dan- 
gers inherent in a university holding patents, MIT 
decided to turn to the Research Corporation of 
New York City. It was to handle all patent matters 
(including determination of patentability and 
allocation of shares to MIT, sponsors, and facul- 
ty) and was to deal with all related legal ques- 

(U) Even the great Vannevar Bush found it 
difficult to raise funds until the second half of the 
1930s. Bush launched upon an almost frantic 
search for combinations of technologies that 
might attract sponsors. Among other attempts of 
the 1930s, he toyed with a machine to identify fin- 
gerprints; he tried to devise a high-speed pneu- 
matic printer; he played with the use of high- 
speed metal tape and wire systems to send secret 
messages; and he tried to find ways to automate 
libraries. But he suffered through many years 
without the kind of financial support that 
Stratton's earlier policies had promised, and, 
most telling, he could not find the financing need- 

ed for what emerged as his grand plan for the 

( U) Bush's Great Plan 

(U) After testing reactions at the Rockefeller 
and Carnegie Foundations, and after considering 
his possible role in Compton's drive to make MIT 
scientifically respectable, Bush put together a 
grand plan. It was one he thought would attract a 
wide range of donors, would be applauded by the 
scientific community, and would lead to a perma- 
nent source of support for the Institute. As well, it 
would call upon the experience and talents of fac- 
ulty from several ofMITs departments. Bush 
decided to make MIT the national center for cal- 
culation and for the development of path-break- 
ing scientific calculation devices. If Bush had his 
way, MIT, not the National Bureau of Standards 
(NBS), would realize Stratton's dream. 

(U) Bush knew that his Analyser had taken 
mechanical technology to its extreme, so his plan 
for the Center of Analysis included much more 
than proposals to extend mechanical analog cal- 
culation. Electronics, photoelectricity, and new 
memory media were to be developed and com- 
bined to produce revolutionary computers. Bush 
also wanted the center to explore the new mar- 
kets for what would later be called "data process- 
ing." His plans included digital calculation and 
machines to solve the escalating problem of file 
management in science and bureaucracy. 11 He 
announced that he would create machines that 
would outdistance all competitors, especially the 
IBM tabulator. 12 Supported by a group of gifted 
junior faculty and a cadre of adoring graduate 
students, hejoined together all ofthe existing 
measurement and calculating projects at the 
Institute and began to weave new ideas for future 
devices. 13 

(U) Beyond Analog Mechanical Machines 

(U) Mechanical analog devices were 
approaching their limits of precision and speed in 


Page 11 


the 1930s. Although there were no commercial 
competitors for such huge devices as the 
Differentia] Analyser, ' 4 Bush saw little worth in 
cloning it in slightly improved form. If support 
was to be found, he had to make a major techno- 
logical leap in analog computing. But there was a 
more fundamental challenge and opportunity for 
the center: the growing demand for digital calcu- 
lation, something MITs machine builders had 
not yet explored. The rise of the social sciences 
was creating a market for digital calculators, and 
even engineers and physical scientists, who had 
been so well served by analog devices for more 
than a century, were tackling problems that called 
for digital methods. Bush also knew of the 
increasing need for high-speed digital calculation 
in the bureaucratic and business worlds. He 
sensed opportunity because there had not been a 
major innovation in large-scale digital machinery 
since Hollerith patented his Tabulator. 

(U) The call for digital processing merged 
with another growing need, information retrieval. 
Business and governmental files had grown to 
unmanageable proportions. The hand, mechani- 
cal, and electromechanical methods of data 
retrieval were not satisfying bureaucratic 
demands. 15 Influential researchers in many sci- 
ences found it increasingly difficult to keep up 
with their areas of interest because of the deluge 
of articles. Bush and many others lobbied for 
projects that would allow scientists and engineers 
to take the lead in the new field of 
Documentation. l6 

(U) Bush decided to concentrate on the 
exploitation of three technologies: photoelectric- 
ity, digital electronics, and film. Although new, 
these technologies were much closer to being 
ready for application than the still delicate mag- 
netic recording. By focusing on the application of 
these technologies to scientific calculation prob- 
lems, Bush hoped to be innovative and to avoid 
conflict with commercial firms. 

(U) Responding to positive reactions by the 
Rockefeller Foundation, Bush sketched a radical 
new design for an Analyser and, by mid-1936, 
succeeded in raising the funds he needed to build 
the next generation of his great analog machine. 
The Rockefeller Differential Analyser was to be 
much faster and much easier to program than the 
mechanical version. Although it remained an 
analog device, it incorporated electronics, digital 
circuits, some photoelectric parts, and program 
tapes. These allowed Bush to eliminate most of 
the cumbersome mechanical components of the 
first model. The new Analyser soon became a very 
demanding, over-budget, and behind-schedule 
drain on the resources of the Institute and a bur- 
den to its students and faculty. The long-delayed 
appearance of the Rockefeller Analyser also 
became a threat to the credibility of Bush and the 
electrical engineering department. 17 However, 
based on the new developments in electronics, 
photoelectricity, and film, he was moving into 
digital calculation and what we now call informa- 
tion retrieval. By the mid-i930s, Bush had rough 
plans for an electronic "programmed" computer 
and refined ideas about information machines. 18 

( U) Two Men with a Need 

(U) A visit by Admiral Stanford C. Hooper and 
his young assistant, Joseph Wenger, would lead 
to one of the most bizarre episodes in American 
history, would complicate Bush's task of estab- 
lishing his center, and would link MITs foray into 
information machines with the world of secrecy. 
The Hooper-Bush agreement for the develop- 
ment of radically new cryptanalytic machines for 
the navy's codebreakers had the potential to set a 
positive role for academic scientists in the inven- 
tion and evaluation of military technology. Its 
promise was not realized, however. The project 
turned into an exercise in bureaucratic bickering. 
More than half a decade was spent dealing with 
organizational problems rather than with the 
technical barriers that were holding back the real- 
ization of the potentials of electronic technology. 
Despite all the efforts of Stanford C. Hooper, 

Page 12 



Joseph Wenger, and Vannevar Bush, the United 
States lost an opportunity to complete the first 
electronic data processing machines and to make 
them operational before the attack on Pearl 

(U) A Man for the Navy 

(U) Stanford C. Hooper prided himself on 
being an innovator, and he devoted his career to 
introducing new technology to a usually reluctant 
United States Navy. Graduating from Annapolis 
in the early 1900s and assigned to the 
Pacific fleet, he immediately began 
to create the navy's first radio 
system. Transferred to 
Washington, he stole hours to 
study at Samuel W. 

Stratton's new National 
Bureau of Standards. 
Mastering the latest radio 
science, Hooper then lob- 
bied for the establishment 
of the navy's own radio 
research division. Hooper's 
expertise and advocacy of 
electronic communications 
soon thrust him into military 
and civilian policy making, 
Although still a young man and a 
junior officer, he was instrumental 
in creating the Radio Corporation of 
America, the giant electronics cor- 
poration formed at government 
request at the close of World War I. 

(U) Because of its need for worldwide com- 
mand and control, the navy had a special stake in 
the success of RCA Hooper hoped that RCA's 
special position would make it confident enough 
to overcome the fear that government work 
would threaten its patents. The hopes of RCA 
serving as a research branch of the navy were not 
completely fulfilled, but Hooper continued to use 
its men and facilities while he searched for help 
from others. 

(U) Stanford C. 

(U) By the early 1930s Hooper was advancing 
through the navy's ranks, was a much-honored 
figure in electronics, and was an acquaintance, if 
not friend, of the leading scientists and inventors 
of the nation. He used such contacts and his 
expertise to devise and forward plans for a fully 
integrated and modern information system for 
the navy, one which was to include every 
advanced technology. He had an even greater 
vision: to permanently wed science and the 
navy. 19 He was determined to prevent the navy 
from being as unprepared as it was for World War 
I. Hooper became tied to those in favor of 
centralized administration and 
increased power for the Chief of 
Naval Operations (CNO). Hooper 
began to develop a strategy, one 
somewhat different from the 
plans of other of the navy's 
new progressive reformers. 
He was willing to depend on 
outsiders. Although he 
helped give birth to the Naval 
Research Laboratory and was 
able to create special research 
sections, such as the Code and 
Signal desk in the Bureau of 
Engineering, he believed the 
navy would have to rely on the 
new research centers that were 
emerging in the largest corporations 
and universities. 

(U) Ending his stay as head of 
the Bureau of Engineering's radio 
section, where he fought for a radio moderniza- 
tion program, Hooper moved from technical to 
more general policy concerns. His appointment 
as Director of Naval Communications in 1928 
gave him an opportunity to aggressively pursue 
his vision. And, when he assumed the newly cre- 
ated position of special scientific advisor for the 
navy in the mid- 1930s and chaired its Technical 
Research Liaison Committee, he had the chance 
to expand his reach well beyond the traditional 
boundaries of communications. All science-relat- 


Page 13 


ed fields, ranging from ballistics through medi- 
cine and atomic energy, became part of his 
domain. 20 

(U) He and his most trusted proteges toured 
the nation seeking ideas and establishing contacts 
with scientists. As part of his plan, he laid the 
bases for permanent cooperation with laborato- 
ries and executives at Eastman, AT&T, General 
Electric, and a host of other corporations. To cre- 
ate a similar link with the universities, he found a 
way to award special military commissions to 
academics so they could remain in the universi- 
ties, yet be a part of the navy's modernization 
effort. 2 * In addition, he collaborated with the 
National Research Council aiding it by finding 
projects and having it help the navy by identifying 
qualified investigators. 

(U) The identification of willing scientists and 
new technologies was only a small part of his task. 
A crucial and politically sensitive step was to con- 
vince the various divisions of the navy to accept 
the civilian men and ideas. The way Hooper han- 
dled that had a great deal of influence on the long- 
term history ofthe automation of American 
cryptanalysis and wedded the history of such 
machines asVannevar Bush's Comparator and 
Selector to the broader straggle for professional 
control within the navy. 

(U) Hooper's admiration for the country's top 
men led him to attempt to force ideas upon 
unwilling navy bureaucrats and skeptical techni- 
cians. Asa result, he alienated many powerful 
men. By 1937 serious complaints reached the 
naval hierarchy about what was seen as interfer- 
ence in the affairs ofthe various bureaus. Hooper 
had to defend himself to the Chief of Naval 
Operations. After the confrontations and the 
complaints to the CNO, Hooper softened his 
approach, but he continued to advocate the types 
of technological innovations that did not fit with 
the service's existing bureaucratic structure. He 
went ahead with his effort to modernize and pro- 

fessionalize the navy, but the political battles of 
1936 and 1937 took their toll on him. 

(U) Even when ill health and perhaps some 
political complications arising from his worries 
about America's military readiness 22 led to a 
reduction of his efforts in the 1940s, he remained 
an important advisor on technical and scientific 
matters and a member of such high science and 
big budget organizations as the National Advisory 
Committee for Aeronautics. By the time he for- 
mally retired in 1943 he had, along with a few 
other senior officers, laid the intellectual if not 
organizational foundations for the Office of Naval 
Research. The ONR became the organization the 
navy successfully used to bring academic science 
into the military after World War II. The ONR 
became one ofthe major sponsors of applied 
mathematics and computers in the United 
States. 23 

(U) Hooper's influence did not end in 1943. 
Although retired, he continued as a consultant to 
major corporations and became deeply involved 
with a company founded by some of his admiring 
young men. The fascinating postwar Engineering 
Research Associates (ERA) was planned as a 
showcase for some of Hooper's dreams. It was to 
be a private company serving the advanced scien- 
tific needs ofthe military. ERA became the torch- 
bearer for the navy's advanced cryptanalytic com- 

(U) Another Plan for Science and the Navy 

(U) Hooper's model for research, which cen- 
tered on cooperation with the private sector, was 
not the only one put forward by navy reformers. 
During the 1930s, one ofthe navy's progressives 
was much less trusting of outsiders. Harold 
Bowen, one ofthe fathers ofthe Office of Naval 
Research, put his energies to strengthening the 
navy's own science and development capabilities. 
While chief ofthe Bureau of Engineering from 
mid- 1935 to 1939, ' M Bowen came in conflict with 
the Bureau of Construction and its allies, the pri- 

Page 14 


T"P fF'*FFT7r'?M IMT " nrl ™nri> inr mi l nnn m i n i HHHVi 

vate shipbuilders. The issue was the design of the 
navy's new destroyers. Bowen was the political 
loser and he remained convinced that the secre- 
tary of the navy's order to merge Engineering and 
Construction into one new agency, the Bureau of 
Ships, was a victory for the technical and political 
Mossbacks. ^ 

(U) Like Hooper, Bowen made many enemies 
because of his fight to keep the navy up to date by 
bringing in new ideas from industry and acade- 
mia. Bowen wanted more research within the 
navy and had faith in a revitalized Naval Research 
Laboratory. One of his last acts as chief of the 
Bureau of Engineering was to create its Office of 
Research and Inventions. With the experienced 
Lybrand Smith and some very enthusiastic young 
officers on its staff, the ORI began to do what 
Hooper bad been advocating for years: integrate 
Engineering with the most talented men in pri- 
vate industrial research laboratories and univer- 
sities. The Office of Research and Inventions 
became the navy's organization to coordinate 
with Bush's NDRC. That led Lybrand Smith and 
Vannevar Bush to become quite close despite the 
growing frictions between Bush and Bowen over 
research policies. Smith also became an impor- 
tant player in the history of OP-20-G's first crypt- 
analytic machines. 

(U) By the end of World War II, Smith and 
Bowen had convinced the navy to create some- 
thing Hooper had always wanted, the Office of 
Naval Research. Bowen made sure the ONR had 
the money, power, and contracting laws to con- 
trol the relationships it established. The ONR 
would use academia and industry to bring science 
to the navy, but it was given enough power to 
allow the navy, not civilians, to direct research. 
Bowen hoped that it also had enough power to 
withstand the protests of the old bureaucrats and 
politicians. 26 One of Bowen's motives for estab- 
lishing the ONR was to allow the navy to develop 
its own program for atomic energy which, he 
hoped, would lead to an atomic-powered ship 

(U) Hooper Confronts the Bureaucracy, 

(U) Stanford Hooper viewed science, 
research, and innovation as significant to every 
naval activity, but he maintained a special inter- 
est and role in naval communications. His plans 
for advancing radio communications led him to 
become involved with the navy's cryptanalytic 
branch, OP-20-G. Hooper became a crusader for 
the expansion and modernization of American 
interception, codebreaking, and all other signals 
intelligence capabilities. It was that involvement 
that eventually led Hooper to MIT in late 1935. 

(U) As with his electronics work, Hooper's 
plans for cryptanalysis came to center on institu- 
tionalized scientific research. At the same time, 
he supported the expansion of the navy's cryptan- 
alytic operating division, OP-20-G. 

(U) For historical reasons, Communications 
(OP-20) rather than the Office of Naval 
Intelligence housed the cryptologic department 
that became known as OP-20-G. 27 And for other 
bewildering reasons, OP-20-G depended upon 
the Bureau of Engineering for the design, pur- 
chasing, and manufacture of its equipment. 
Another naval branch handled contractual 
details. To further complicate the bureaucratic 
tangle, OP-20-G's Research Section (Y) was the 
small group charged with communications secu- 
rity and, significantly, the exploitation of the lack 
of security of the communications of other 
nations. On top of that, and despite OP-20-G-Y's 
mandate, yet another research group was set up 
within Engineering to explore related technologi- 
cal questions. Adding to the confusion over power 
and domain was the Naval Research Laboratory. 
Although their core functions were under the 
direct command of the CNO in the critical years of 
the 1930s, communications and cryptanalysis 
had a tough go of it in the navy. 


Page 15 


(U) A Few Men and Women for Secrecy 

(U) The navy's very small ciyptologic group, 
OP-20-G, began its life during World War I but 
was not active until the mid-i920s. 28 One reason 
for its inaction was that just as it was founded, the 
incredible Herbert Yardley was lobbying for the 
creation of what became the famous American 
Black Chamber. His group was to serve the crypt- 
analytic needs of the army and the State 
Department, and, to some unknown degree, the 
navy. Stealing resources away from the private 
cryptanalytic group that had been developing at 
the estate of the flamboyant millionaire, Colonel 
Fabyan, Yardley achieved some amazing victo- 
ries. He broke the codes and ciphers of the major 
powers. That allowed the United States to predict 
the bargaining positions of the important players 
in the naval arms limitation negotiations of the 
1920s. Yardley's work made him some good 
friends but also some enemies. A few rash deci- 
sions on his part also led to the closing of his 
Chamber in 1929 and the transfer of its files to the 
army's old code organization under William 
Friedman. Yardley then decided to take one of the 
most fateful steps in the history of American 
cryptanalysis. He published a book that told the 
when, what, and why of American cryptanalytic 
success. One horrible consequence was that the 
Japanese began to change all their code and 
cipher systems. 29 

(U) OP-20-G did not receive much official 
navy support. Until the mid-i920s, when it came 
under the command of a young and bright officer, 
Laurance F. Safford, it was almost a shadow 
organization. Safford arrived just in time to take 
advantage of the "acquisition"' of a copy of part of 
Japan's secret naval code. The code proved 
invaluable, and OP-20-G began providing critical 
information to the navy. But that did not mean 
recognition of the potentials of communications 
intelligence or adequate funding. "G" might not 
have survived if it had not been for a supersecret 
fund set up at the end of World War I. 

(U) OP-20-G's interactions with the Office of 
Naval Intelligence were, at times, ones of strain as 
well as frustrating dependency. ONI did much of 
the needed dirty work to obtain codebooks and 
information about cipher machines, 30 and it had 
the responsibility for interpreting the intentions 
of America's enemies. But the ONI and OP-20-G 
were bureaucratically separate, and at key times 
there was mistrust. "G" also had less than satis- 
factory relations with naval commanders. The use 
of"G's" information was dependent upon the 
decisions of local commanders, and OP-20-G 
relied upon their willingness to supply intercepts 
to Washington. Even serving only a technical 
cryptanalytic role was difficult for "G." It took 
many years for it to acquire any control of what 
radio systems were to be monitored. 

(U) Captain 

Laurance F. 



(U) The Search for Pure Cryptanalysis 

(U) Through his academic and corporate con- 
tacts, Hooper learned of the potentials of mecha- 
nized automatic control and of the increasingly 
mathematical nature of science and cryptanaly- 
sis. His awareness of the expanding reach of sta- 
tistical techniques, the potentials of high-speed 

Page 16 

■mo ccrpETHmwimiPEi ™ iica , amc r bm , rati Aim hti » 


calculators, and the use of light-sensitive devices 
in astronomy were perhaps sharpened by visits 
and discussions with Vannevar Bush. 3 ' Whatever 
the particular source of his knowledge, Hooper 
believed that the new electric and mechanical 
ciphering devices introduced by the major pow- 
ers, including the United States, would force 
cryptanalysts to become statisticians. They would 
have to perform seemingly impossible feats of 
calculation to penetrate the ciphers produced by 
such complex machines as the Kryha and the 
Enigma. 32 

(U) As soon as he assumed command over 
Communications, OP-20-G informed Hooper of 
its progress against the cipher machines. The 
cryptanalysts were quite proud of their secret and 
clever techniques, ones they thought were essen- 
tial because of the impracticality of a pure mathe- 
matical approach. 33 Although they employed sta- 
tistical techniques, they had effective short cuts 
such as finding a copy of a secret message sent in 
a known code; locating often repeated phrases 
(cribs); or uncovering the pattern of the way an 
enemy announced the wheel settings for a cipher 
machine network. They were also quite proud of 
their craftsmen's tools, such as paper wheels, long 
strips of wood with alphabets painted on them, 
and overlay sheets with punched holes for attack- 
ing ciphers. But Hooper and his new right-hand 
man, Joseph Wenger, were not impressed by the 
tricks, and they thought that OP-20-G's technolo- 
gy, if not methods, were woefully behind the 
times. In late 1930 Hooper suggested to OP-20-G 
and the Bureau of Engineering that they begin to 
develop automated cryptanalytic machines and, 
by implication, to formalize their approach to 
analysis. 34 

(U) Hooper wanted machines that would free 
OP-20-G from tricks and dependencies and that 
would allow the use of advanced mathematics. 
Those machines would have to be innovative 
because the new cipher devices presented crypt- 
analysts with problems far different from those of 
code systems. Codes were secret lists of words (or 

(combinations of numbers) that stood for other 
words. In contrast, cipher machines dynamically 
changed letters into different ones with no pre- 
dictable relationship between the original and the 
cipher letters. The limited vocabulary of a code 
meant that acquiring a copy of its codebook was 
an effective solution, unlike the situation with 
sophisticated cipher machines in which having a 
copy of the enemy's machine was only a small 
step toward reading messages. The key method 
an analyst used to solve a code was to identify the 
relationships between a particular code word and 
other words. Correlation analysis and the use of a 
decoded word to predict the meaning of another 
were viable methods. 

(U) The new cipher systems demanded less 
obvious approaches. The cipher system design- 
ers' goal during the 1920s and 1930s was to avoid 
the meaning embedded in any code system. The 
American Hebern cipher machine and its 
European cousins, such as Enigma, took the old 
principle of random substitution of one letter for 
another to a new level. They went far beyond the 
centuries-old cipher tables and handy substitu- 
tion algorithms. 

(U) All of the new machines relied upon sets 
of wired rotors (or relay analogs of them) whose 
internal electrical connections produced a unique 
substitution cycle of such complexity and length 
that it could be penetrated only through time con- 
suming analysis of forbidding amounts of data. 
Unless the operators of the encryption machines 
made a mistake, or the cipher breakers had a con- 
stant source of information on the settings of the 
cipher wheels, incredible amounts of calculation 
were needed for pure cryptanalysis. Hooper was 
sure that the growing use of the new cipher 
machines and the shortage of experienced crypt- 
analysts meant an end to the power of informal 
methods. He saw no alternative but to develop 
formal techniques and advanced machines. 

(U) More than an abstract faith in scientific 
cryptanalysis led to Hooper's drive for new 


Pige 17 


machines. There were very practical reasons. "G" 
had to be made independent and ready for an 
emergency. Older methods, for either codes or 
ciphers, demanded too many experienced code- 
breakers who had spent years working on partic- 
ular systems and on information supplied to OP- 
20-G by others such as Naval Intelligence. 
Automation and formal procedures would have to 
substitute for professional skill and experience as 
well as the old codebreaker's standby, intuition. 

(U) But in 1930 the navy's bureaucracy and 
even the crew at OP-20-G were less than accept- 
ing of formal analysis and machinery. The code- 
breakers atOP-20-G were aware of the emer- 
gence of the new ciphering devices and, in fact, 
were building their own versions as well as tack- 
ling the systems of other nations. Because of their 
direct experience with automatic enciphering 
devices, Hooper's September 1930 "suggestion" 
about methods and automation was not too well 
received. OP-20-G's principal civilian cryptana- 
lyst, Agnes Meyer Driscoll, did not like the idea at 
all. Additionally, the cryptanalysts felt insulted 
because Hooper's request contained an implicit 
criticism of their work and skills. They thought 
that formal methods, while helpful, would never 
replace an experienced codebreaker. And their 
years of work had taught them that decryption 
was usually dependent upon some type of infor- 
mal initial entry into a system, whether it be a 
psychological insight, a theft of materials, or the 
transmission of a message in both clear and enci- 
phered form. 35 In addition to the codebreakers' 
distrust of those who proposed unrealistic meth- 
ods and machines, the small OP-20-G staff was 
too busy analyzing Japanese code systems to deal 
with methodological speculations. 36 

(U) Hooper thought he would eventually 
tempt OP-20-G into applying formal methods by 
presenting it with a demonstration device. 
Hooper soon arranged to have the Bureau of 
Engineering create anew section for advanced 
code and signal research 37 and then made sure 
that someone who would pursue his goals filled 

the post. A young officer who had been one of the 
first students in OP-20-G, who had experience as 
a seagoing communications officer, and who was 
already a protege of Hooper, was selected. Joseph 
Wenger, a thirty-year-old 
Annapolis graduate, 
followed Hooper's 
cues and began a 
search for new 
for all aspects 
of communi- 
cations with, 
of course, an 
eye open for 
new devices 
for ciphering 
and decipher- 
ing messages. 
With some inter- 
ruptions caused by 
shifting naval assign- 
ments, Wenger con- 
tinued that search 
through the 1930s 
and 1940s, and he became the driving force 
behind what became the most technically 
advanced cryptanalytic agency in the world by the 
late 1940s. 38 

(U) From Electronics to Electromechanics 

(U) In the early 1930s Hooper's academic 
contacts turned him towards something much 
more innovative, the electromechanical tabulat- 
ing machines built by companies such as IBM and 
Powers. Hooper successfully prodded the Chief of 
Naval Operations into sending a very specific and 
strong directive to the Bureau of Engineering in 
late 1931. 39 It ordered the Bureau to devote 
resources to study the new optical sorters and 
special devices for blind reading and came close 
to demanding that such technologies be used to 
build a deciphering device. 40 The CNO's mandate 
included more than cryptanalytic investigations; 
it was a signal to Hooper to intensify his efforts to 

(U) Joseph 

Page 18 



link science to the navy. Under pressure from 
Hooper, the Bureau provided Wenger with the 
money needed to make a grand tour of America's 
research laboratories. During his visits, Wenger 
encountered fantastic new technologies that had 
at least long-term promise for solving the difficult 
cryptanalytic problems, but most seemed to 
demand a protracted and expensive development 
period. 41 Wenger was especially disappointed 
when he realized that optics and electronics were 
not quite ready to produce a cryptanalytic 

(U) Perhaps because of that and because of a 
sudden realization by OP-20-G that it would need 
some type of mechanical aids, Wenger turned his 
attention to a more established technology. The 
Hollerith and Powers electromechanical tabulat- 
ing and sorting machines were evolving into quite 
sophisticated devices by the late 1920s. In addi- 
tion, they were machines that were immediately 
available for use and were commercially pro- 
duced. Wenger examined the Remington-Rand 
Powers tabulators used by OP-20-G in 1932 and 
did enough research to allow Hooper to again, but 
more authoritatively, suggest that OP-20-G 
investigate them. It was difficult for the officer in 
charge ofOP-20-G, Laurance Safford, to ignore 
Hooper's urging any longer. 42 But Hooper's grand 
dream suffered a temporary yet important set- 

(U) Just as Wenger was exploring the various 
technical possibilities, it was discovered that the 
Japanese had replaced their Red Code with a 
completely revised set that could not be penetrat- 
ed. Perhaps because ofYardley*s indiscretions, 
seven years of work on the previous code had 
become valueless! OP-20-G's codebreakers knew 
they would be unlikely to obtain a copy of the 
Blue Japanese code and the three other new sys- 
tems * 3 and decided to take on the formidable 
task of breaking the code through pure methods. 
The Japanese continued to use the old type of 
superencipherment, the modular addition of ran- 
dom numbers to the code groups, so it was a rel- 

atively easy target. But the code itself, Wenger 
knew, would demand years of work. Over 
100,000 words had to be decoded. Such an effort 
called for either vastly increased manpower or 
mechanical aides. 44 Everyone knew that "G" was 
unlikely to be allocated more men. 

(U) In early 1932 OP-20-G's cryptanalysts 
studied Wenger's tabulator survey and decided to 
select the type centered upon electrical rather 
than mechanical reading of cards. Seeing 
Remington-Rand's system as inflexible, they hur- 
ried to rent the electromechanical IBM tabulating 
devices, ones built to handle alphabetic charac- 
ters as well as numbers. The punch card era 
seemed to have begun at OP-20-G. 

(U) Then the navy hierarchy declared that it 
was unwilling to fund the experiment! Safford 
and Wenger did not give up. OP-20-G pressured 
the Bureau of Engineering to scrape some funds 
from its already slim budget, 45 but the Bureau 
was able to raise only a few hundred dollars, not 
several thousand, to start the project. It contin- 
ued to piece together small amounts during the 
1930s to support the tabulators. But it always felt 
that OP-20-G did not fully appreciate its efforts. 46 
Only a machine or two arrived at OP-20-G, and 
their experimental use, which soon turned into a 
necessity in the eyes of many atOP-20-G, sur- 
vived only as a near underground activity. 

(U) Despite the hand-to-mouth funding of its 
few machines, the OP-20-G tabulator crew con- 
tinued with its work and made major contribu- 
tions to the penetration of the new Japanese 
codes. The navy also explored new ways to store 
data on IBM cards, and during the war it helped 
develop special tabulating machines. 

(U) Ironically, OP-20-G's early 1930s tabu- 
lator-related achievements had a negative influ- 
ence. Although the search for cryptanalytic tech- 
nology and methods had been motivated by 
Hooper's deep fears concerning the new automat- 


Page 19 

T Or S E CWHW O MIN T /m E L TO UflA, AU3, CAM, ODR and N ZUttCI 

ic ciphering machines, including Britain's, 47 the 
crisis caused by the change in Japan's older code 
system shifted attention to more immediate prob- 
lems and forced a commitment to available 
devices. The more sophisticated machine options 
were dropped in favor of the tabulators. The tab- 
ulators were well suited to many decoding proce- 
dures, especially those calling for sorting and, 
later, collating operations, but they were not the 
mathematical or truly high-speed statistical 
devices needed to break into the new cipher 

(U) The leasing of a few tabulators did not link 
IBM to any long-term commitments to OP-20-G 
or Engineering. Although IBM played a signifi- 
cant role in certain extensions of electromechani- 
cal technology before and during World War II, it 
did little truly far-ranging research for the crypt- 
analysts during the 1930s. While the use of tabu- 
lators was a great step in the history of crypt- 
analysis, the commitment to tabulators took away 
much of the incentive to make the great techno- 
logical leap Hooper had desired. The very hard- 
pressed staff at OP-20-G had more than enough 
to do to learn how to exploit the IBM equipment. 

(U) Then, when older cryptanalytic methods 
triumphed over Japan's new cipher machine, the 
Red, there was little excuse for an emergency 
development program. The success against Red 
undermined arguments that an advanced in- 
house developmental group should be estab- 
lished within the Bureau. 

(U) A Young Man for the Future 

(U) Something else helped to turn the navy 
away from Hooper's plans for truly advanced 
automated cryptanalysis. Joseph Wenger, 
Hooper's man in the Bureau or Engineering, who 
had become an ardent believer in the value of sci- 
ence and technology, was returned to sea duty in 
mid-1932. He had supplied Hooper's grand out- 
line for communications with the details needed 
for OP-20-G's technical and organizational 

future. 48 As significant, while on sea duty, he 
refined and codified the important method later 
known as traffic analysis (T/A). He combined 
direction finding, callsigns, and traffic flows into 
a highly effective tool. 49 To prove the worth of the 
approach, he reconstructed the Japanese naval 
maneuvers without being able to read the con- 
tents of the radio signals. 50 Although not appreci- 
ated by outsiders, even Hooper during the early 
1930s, T/A became a major factor in the 
American victory in World War II. 

(U) The Dream Postponed, Again 

(U) Wenger's transfer to sea duty in 1932 
allowed him to help unravel Japan's naval tactics 
and to refine America's eavesdropping capability 
in the Pacific, but it was near devastating to the 
cause of automating code and cipher breaking. 
Almost as bad for Hooper's cause was Laurance 
Safford's assignment to sea for four years. His 
absence until 1936 stretched the resources of OP- 
20-G to the breaking point and left Hooper with- 
out an in-house advocate. When Safford returned 
to Washington, the growing crisis in the Pacific, 
including the sudden change of a major Japanese 
code in 1936, left him with no time for experi- 
mentation. Despite OP-20-G's dependency on the 
Bureau of Engineering for hardware develop- 
ment, the engineering branch was left without a 
spokesman for advanced cryptographic technolo- 
gy. What men Engineering could spare became 
involved in the difficulties of inventing and man- 
ufacturing electromechanical encryption devices. 
The bureau, along with the Naval Research 
Laboratory, also faced increasing demands and 
few thanks for radio and radar developments. At 
the same time, OP-20-G became deluged with 
new and more difficult code and cipher problems 
as Japan carved out its Asian empire. The tiny 
crew had little time for technological or mathe- 
matical speculation. 

Page 20 


T O P S E eR ET //C O M I N T // f tEL TO USA AU3, CA N , QDR A N D N ZUOd 

(U) The Dream Reborn, for a Moment 

(U) It was only Wenger's return in mid-1935 
and the Roosevelt-Vinson decisions to expand the 
navy that allowed Hooper to again pursue his 
cryptanalytic goals. Wenger had the experience, 
the energy, and the desire to restart the program, 
and naval expansion hinted at the possibility of 
funding. 52 

(U) The changes at OP-20-G in 1935 extended 
to more than the renewed hopes for new research 
machinery. Wenger was made the head of OP-20- 
G's new research desk. The new 'Y'section was to 
be devoted to the application of science to crypt- 
analysis and to the type of long-term planning 
development that the CNO was encouraging in all 
parts of the service. 53 Then, when Safford came 
back to Washington in 1936, Wenger began 
another round of visits to the centers of American 
science and technology. Among those Hooper 
visited in 1935 and then recommended to Wenger 
was a man he had known for years, Vannevar 
Bush. 54 

(U) Bush, Wenger, and Hooper joined forces 
at a time when their interests seemed to be in per- 
fect harmony and when they thought they had the 
resources and power to initiate and complete a 
major program. Bush's scientific status was per- 
haps the major reason why Hooper looked to MIT 
rather than to the large corporations such as 
National Cash Register or IBM or RCA or to the 
National Bureau of Standards for help in 
automating American cryptanalysis. On a gentle- 
men's agreement, Bush began to draft a plan for 
the navy, and Wenger returned to Washington 
filled with enthusiasm. He was convinced that the 
$10,000 consulting fee Bush expected was a great 
bargain. Bush dashed off his report and submit- 
ted it in the first weeks of 1936. He was able to 
respond so quickly because of the optics-film- 
electronics work he and his colleagues at MIT had 
been doing for several years. Of great importance, 
he had begun thinking of and lobbying for the 

development of electronic cryptanalysis well 
before 1935- 55 

(U) Bush's initial proposition was not for the 
production of specific equipment. Rather, it 
defined his role as that of a consultant to the navy. 
He sketched the general outlines for a long-term 
project centered about the creation of high-speed 
optical-electronic devices which would be hun- 
dreds of times more powerful than the tabulators. 
He recommended that the navy design and devel- 
op what became known in the intelligence com- 
munity as Rapid Analytical Machines (RAM). 
Everything finally seemed to be falling into place 
for them and Bush in early 1936. 

(U) Little Science Meets the Little Navy, 

(U) Hooper thought he was having Bush sub- 
sidize his great plan for the navy. Bush thought 
the navy would subsidize the beginnings of MITs 
calculation center and its entry into digital pro- 
cessing. Wenger thought he had a set of ideas that 
would launch the navy on a full-scale develop- 
ment project. None of them realized there were 
built-in conflicts. Hooper probably did not know 
of the financial pressures on Bush and MIT dur- 
ing the 1930s. In turn, Bush did not suspect that 
Hooper and Wenger had not convinced the navy 
of the worth of their approach to introducing 
innovations. 56 

(U) Just as the prospects for Bush's center 
rapidly brightened and as Hooper was receiving 
signals that his comprehensive plan for all com- 
munications activities would be approved, the 
navy made an unexpected, critical, and disap- 
pointing decision. For a second time the attempt 
to revolutionize cryptanalysis seemed to have 
been defeated by the tangled navy bureaucracy 
and the men Bowen called "mossbacks"! Before 
Bush's navy project truly got under way, he and 
his naval allies became involved in an organiza- 
tional nightmare. Bush thought freedom from 
interference was essential if academia and the 


Page 21 


military were to join together, and he believed 
that no absolute timetables and guarantees could 
be given for truly innovative work. Hooper and 
Wenger agreed that heavy-handed bureaucratic 
oversight would doom any creative effort. 
Wenger hoped that Bush's status and persuasive 
powers would be able to break the navy bureaus' 
resistance to outsiders. But the naval bureaucracy 
had a different opinion. 

(U) The Bureau of Engineering men very 
bluntly told Hooper and Wenger that Bush's 
plans were unrealistic and his demands outra- 
geous. They were soon joined by the contracting 
arm of the navy, which declared many parts of 
Hooper's model for academic/military coopera- 
tion ill advised, if not illegal. They would not give 
the needed approval, and the project that could 
have led to the creation of the first electronic dig- 
ital data processing device seemed dead in early 

(U) A Man for Statistics 

(U) Just as Stanford Hooper was facing the 
defeat of his hopes of creating anew technology 
for cryptanalysis, another major figure in the his- 
tory of American codebreaking was becoming 
entangled with automation. William F. Friedman, 
the head of the army's cryptologic section, finally 
convinced the army to allow him to use tabula- 
tors. Although their introduction into the army's 
Signal Intelligence Service (SIS) came almost five 
years after Hooper and Wenger had brought 
them into OP-20-G, the arrival of the IBM 
machines at the SIS offices seemed revolutionary. 

(U) In 1929-30, just as Hooper was trying to 
refurbish naval communications, the army had to 
fill the void left by the disintegration ofYardley's 
Black Chamber. Instead of creating an entirely 
new organization, it gave additional mandates 
and some additional resources to the man it had 
previously hired to safeguard its own communi- 
cations, William F. Friedman. Unlike Wenger or 
Hooper, Friedman had not come to code work 

through the military; rather, he stumbled into it 
because of his college courses in genetics. 

(U) The son of a Hungarian-Russian-Jewish 
immigrant, Friedman attended an advanced 
technical high school where he delved into elec- 
trical engineering. But his interest in the new field 
of scientific agriculture led him to one of 
America's centers of applied science, Cornell 
University. After finishing heavily statistical 
courses in genetics, and gaining experience in 
research at one of the pres- 
tigious Carnegie cen- 
ters, Friedman 
decided to post- 
pone gaining a 
Ph.D. He want- 
ed and needed 
a job. He 
accepted a 
position as a 
for one of 
America's most 
influential agri- 
cultural business- 
men, Colonel George 
Fabyan. Assigned to 
Fabyan's estate at 
Riverbank, Illinois, 
just as Europe was becoming engulfed in war, 
Friedman soon found himself busy with Fabyan's 
private cryptanalytic projects rather than with the 
development of hybrid cottons. When Fabyan 
offered his staff and his estate to the United States 
government for cryptanalytic training for the war 
effort, Friedman's future was set. 

(U) His energies were turned to applying the 
new statistical techniques he learned at Cornell to 
cryptanalysis. His cryptoattacks and his training 
methods became legendary. As a result, after the 
war the United States Army asked him to estab- 
lish a code agency. Because Herbert Yardley's 
Black Chamber held the mandate for listening to 

(U) WiHian F. Friechan 

Page 22 


TO P S e eR ET //eOMIMT/mgL TO U3A, AU3, CAM, QDR A N D NZU/X 1 

the communications of others, Friedman was 
asked to focus on the protection of army commu- 
nications and on the preparation of training man- 
uals for wartime cryptanalysis. Although concen- 
trating on those tasks, Friedman did not abandon 
code and cipher breaking. He was called on to test 
various proposed systems, including cipher 
machines the navy thought of purchasing. 57 

(U) Friedman's role began to change in 1929 
when Yardley's group was under political threat. 
The army decided to found its own operational 
cryptanalytic group. It gave Friedman the funds 
he needed to hire a group of young civilians, and 
it gave him Yardley's files. Perhaps it gave him 
access to Yardley's old sources of intercepts. 
Friedman trained his young men in codebreaking 
and made sure they learned about formal statis- 
tics and foreign languages by enrolling them at a 
local university. Meanwhile, his wife, also a 
Riverbank alumna, became the cryptologist for 
the Coast Guard. 58 While her crew worked on the 
clandestine messages of rum runners and other 
criminals, Friedman's team examined as much 
diplomatic and military traffic as it could obtain 
through the very limited intercept capabilities of 
the army. 59 

(U) Together, the Friedmans blended practi- 
cal experience with statistics to develop more 
powerful cryptanalytic tools. Although Friedman 
did not attempt to make the direct links with elite 
academics that Hooper was forging for OP-20-G, 
he was proud of the "scientific" character of his 

(U) Those statistical methods and knowledge 
of many of the machine activities at"G" soon led 
Friedman to seek a means of automating the 
army's codemaking and codebreaking work. 
Beginning a few years later than the navy, 
Friedman tried to acquire IBM tabulators for his 
office. He faced almost as many frustrations as 
Hooper and Wenger but finally acquired some 
machines in late 1934. 60 By 1937 he and his crew 
had developed several tabulator methods that 

became classic means of cryptanalytic attack, and 
they began to turn the tabulators into more spe- 
cialized cryptologic tools. He and one of his young 
men, Frank Rowlett, invented an attachment for 
the tabulating equipment that allowed it to gener- 
ate "random" code. 61 

(U) Friedman began to develop visions of a 
greater technological future for cryptanalysis. 
But, unlike Hooper and Wenger, he did not seek 
help from outsiders, at least not in the 1930s. 
Perhaps that was because his research ambitions, 
even more than "G's," were smothered by the mil- 
itary bureaucracy. Friedman did not have a 
Hooper to ran interference for him with the 
Signal Corps. For whatever the reasons, 
Friedman's automation efforts were less adven- 
turous and more limited than Wengefs. He had 
noVannevar Bush and no ties to the nation's sci- 
entific elite. 

(U) In the mid-i930s Friedman concentrated 
on plans for putting teletype tape readers, relays, 
and plugboards together in various combina- 
tions. Some of those became outlines of his own 
versions of Index of Coincidence machines and 
isomorph locators (pattern finders). 62 And, at the 
end of the decade, he somehow found the money 
to hire an MIT electrical engineer, Leo Rosen. 
Rosen had a solid background in electronic tubes 
and circuits. Perhaps Friedman hired him with an 
eye to beginning his own version of the navy's 
electronic RAM program. 

( U) Science and the Navy Need Other 

(U) Tn early twentieth century America, cor- 
porations and private foundations were more 
important than government or higher education. 
As a result, corporate research policies and deci- 
sions by the leaders of the philanthropic founda- 
tions played a determining role in the history of 
Bush's and Hooper's crusades. Decisions by 
Eastman-Kodak, AT&T, IBM, and especially the 
National Cash Register Company were critical to 


Page 23 


the emergence of the machines for cryptanalysis 
and for the library. As late as the 1930s, the sci- 
entists' lobbying efforts to make pure science one 
of the targets of federal support were failures. 
They were rebuffed by Congress as well as by the 
usually open-handed Franklin D. Roosevelt. As a 
result, there was no pure science program in the 

(U) The Private World of Science 

(U) During the first forty years of the twenti- 
eth century, the nation's scientists looked to two 
sets of foundations, those created by Andrew 
Carnegie and John D. Rockefeller. Their fortunes, 
generated by the technological revolution of the 
nineteenth century, became the fuel for American 
academic science. 

(U) Their decision in the 1920s to finance 
research within the elite American universities 
was critical to the history of American science. As 
important, they created the first bureaucracies 
designed to manage long-term, very expensive 
scientific programs. Those programs accounted 
for perhaps as much as 90 percent of such activi- 
ty during the 1920s and 1930s, and their man- 
agers became key players in the shaping of scien- 
tific institutions during and after World War IL 63 
The administrators of the 1930s private founda- 
tions, including Vannevar Bush, became the over- 
lords of 1940s science and then became the lead- 
ers of the early Cold War scientific and high-tech 

(U) As outside research became more attrac- 
tive, the Carnegie and Rockefeller foundations 
turned to the old national science institutions for 
help. The National Academy of Sciences and the 
National Research Council were energized with 
foundation monies and began to act as scientific 
go-betweens. The NRC managed many projects 
for Carnegie, advised other foundations about 
national needs, and recommended worthy scien- 
tists. 64 After those first steps, the foundations 
began to help some individual academic 

researchers just as MITs new president launched 
his faculty, including Vannevar Bush, on a sweep 
for research funds. Very important to Bush were 
the decisions by one of the new young adminis- 
trators at the Rockefeller Foundation, Warren 

(U) A Man for Applied Mathematics and 

(U) Warren Weaver was one of those new 
bright scientific men brought to the foundations 
to reformulate policy. Central to Weaver's plan 
for the revamped natural science division of the 
Rockefeller Foundation was the creation of 
instruments to encourage the use of mathematics 
in every field. By the mid-i930s Bush convinced 
Weaver that the world of science was ready for 
new generations of Analysers. Then Weaver suc- 
cessfully lobbied his superiors for aSio,ooo 
study grant for Bush's proposed partially elec- 
tronic machine. Just a year later, he secured an 
astounding 885,000 for the Rockefeller Analyser 
project atMlT. 63 Haifa dozen years later, Weaver 
again showed his faith in MIT when he funded 
another huge computer project at the Institute, 
one for an electronic digital programmed com- 


(U) American Science and the War - the 

(U) Only a few in America realized that 
Germany was inventing anew type of high tech- 
nology warfare and that fundamental science 
might be needed to combat the horrors of atomic 
weapons and long-range bombers. Vannevar 
Bush and his close scientific friends were among 
those few. Never a man to sit by and let the world 
determine his fate, Bush sought ways to ensure a 
flow of academic contributions to the war effort. 67 
Bush energized what became one of America's 
first modern science interest groups and began to 
lobby the government to support a wide range of 
new programs. Bush convinced President 
Roosevelt to create the powerful and well-funded 

Page 24 


top cccncT/;oo Mi NT//ni:L to uoa, auo, can, obr an p hzu/xi 

National Defense Research Committee (NDRC) 
in June 1940. Within a year, its scope and its pow- 
ers to initiate and control projects were vastly 
expanded. The new Office of Scientific Research 
and Development was a dream come true for 
Bush. It was almost the perfect science founda- 
tion for elite American academics. 

(U) The NDRC was responsible to the presi- 
dent, not the military or Congress, and its scien- 
tists could determine what projects to begin or 
end. Hundreds of millions of dollars came under 
the control of the NDRC. The NDRC and elite sci- 
ence were subsidizing science as well as potential 
weapons. Administrators of foundation science, 
who were friends of the universities, were select- 
ed to head the major branches of the NDRC. The 
old Carnegie-Rockefeller circle, which included 
the leading men from the leading universities, 
moved from private to military philanthropy dur- 
ing the war and, along with Bush, were able to cir- 
cumvent the "mossbacks" in the military and the 
older organizations of science. In the fall of 1940, 
the NDRC began to explore defense technologies 
that were too speculative for the military or its 
older industrial allies. Of great importance was 
the computer effort headed by Warren Weaver. 

(U) Because of Weaver's mathematical back- 
ground and his prewar experience evaluating 
computing proposals, Bush made him head of the 
mathematical and scientific instrument section of 
the NDRC. One of his first chores was to develop 
a program to solve technical problems created by 
the advance of German military technology. 
There was a vital need for automatic control of 
antiaircraft weapons, high-speed counters for 
ballistic tests, and scientific instalments to moni- 
tor atomic processes. 68 In each case Weaver 
turned to electronic solutions. He called upon all 
those known to have worked in electronic count- 
ing and launched a program for the development 
of special purpose devices. He soon had the com- 
puter builders George Stibitz and Sam Caldwell to 
help him supervise the work. As important, he 
was able to pursue another opportunity. He cre- 

ated a center for applied mathematics. It would 
permanently change academic mathematics in 

(U) The NDRC was a blessing to Bush and his 
academic friends, but to others it was a politicized 
and unnecessary organization that threatened the 
military research agencies such as the Naval 
Research Laboratory. 69 To Admiral Bowen, Bush 
was leading a group bent on playing favorites 
among the military services and the universities. 
He soon concluded that the NDRC worked to the 
disadvantage of the navy. To Admiral Hooper, 
however, Bush and the NDRC appeared, at least 
at the beginning, to be the only way the intelli- 
gence community could acquire the advanced 
machines it needed. But computers were far 
down on the NDRC list, and cryptanalysis entered 
its world only because of the long chain of associ- 
ations between Bush, the navy, and the corpora- 
tions and universities that were at the center of 
the NDRC. 

(U) Corporate Charity 

(U) Vannevar Bush looked to the major cor- 
porations when he began his search for support 
for his calculation center in the early 1930s. 
General Electric had a research branch that was a 
leader in applied mathematics, but it decided to 
keep most of its work in-house rather than make 
any large investments in Bush's center. 
Paralleling General Electric's reaction, Western 
Electric and Bell Laboratories were willing to sup- 
ply critical parts for the Rockefeller Analyser and 
to give advice on the type of tools and services 
mathematicians desired. But they did not offer 
major financial support to Bush's 1930s projects. 

(U) The Eastman-Kodak Corporation of the 
1930s was not as generous with MIT as its 
founder had been, but it remained a very good 
friend of the Institute. Of even greater importance 
to the nation was Bush's relationship with a cor- 
poration that did not have a reputation for 
research. Why Bush became so close to a compa- 

top cr:om:T//oo M iNTOncL to uoa, auo, oa n , odr and n zu/xi 

Page 25 


ny that made cash registers is explained by Bush's 
friendship with the famous team of Colonel 
Edward A. Deeds and Charles Boss Kettering. 
That friendship linked National Cash Register, 
MIT, the NDRC, and the Ultra secret. 70 

(U) Bush first came in contact with Deeds and 
Kettering through the institutions of American 
science. Bush and Deeds served on important 
advisory committees that steered aeronautical 
research in the United States, such as the precur- 
sor to NASA, the National Advisory Committee 
for Aeronautics. In 1931 Deeds consented to serve 
as chairman of the board of National Cash 
Register (NCR). 

(U) For someone trying to rebuild NCR, the 
best opportunities were those that demanded a 
technology not found in the corporation's offer- 
ings of the 1920s. NCR needed new machines to 
move deeper into information processing. 
Inventory, retail sales, and personnel manage- 
ment, for example, had demands met only by 
devices that had some sort of large-scale memory. 
The failure to create offerings to compete with 
IBM was one reason for the demand for a thor- 
ough shake-up at National Cash Register in 1936. 

(U) While Deeds slashed expenditures in 
many parts of the company, he increased alloca- 
tions for research. He pushed the efforts to move 
NCR into the electrified bank-posting and billing 
machine business, and he looked forward to find- 
ing a technology to challenge IBM's grip on auto- 
mated file management. Previously, Deeds had 
applauded NCR's very quiet acquisition of the 
rights to a fantastic machine for the era, the 
Hofgaard relay computer. 71 Both the 1930 and 
1938 NCR relay computer patent applications 
cited a machine with an architecture quite like 
that of the modern serial computer. It had a cen- 
tral processing unit and addressed storage. It per- 
formed at least three of the four basic arithmetic 
functions and had the ability to calculate, store, 
and print totals and subtotals for many different 
items. Although Hofgaard's machine was quite 

promising, Deeds ordered NCR's research direc- 
tor, Harry N. Williams, to drop the project and 
investigate other technological options. Deeds 
was probably advised to do so by Vannevar Bush, 
who was aware of the Hofgaard patents and who 
had just completed his survey of computing tech- 
nologies. 72 Bush advised a jump into electronics. 
The men at NCR learned much about the 
progress of electronics and film-optical combina- 
tions in scientific measurement from Bush. They 
were certainly interested in the MIT work on 
smaller and more reliable tubes because of the 
value of low power and fast miniature tubes for 
machine design. 73 Their positive reaction to the 
operation at MIT resulted in an endorsement of 
Bush's suggestion to use the Institute as a 
resource for NCR. 

(U) The Navy Comes in Second 

(U) After all the disappointing appeals to the 
foundations and the troubled negotiations with 
the navy, Bush finally gained a pliable and gener- 
ous sponsor. Bush turned to Deeds requesting 
money for the proposed universal electronic com- 
puter, the revolutionary Rapid Arithmetical 
Machine. Explaining that it was still on paper, but 
underscoring that other work had already led to 
the building of successful electronic circuits, Bush 
was able to get Deeds's attention. 74 The first dis- 
cussion about the electronic computer may have 
started with hints that MIT could immediately 
build an electronic calculator for NCR But the 
beleaguered Rockefeller project led Caldwell and 
Bush to scale down their ambitions. Bush, already 
very busy, had a limited role in the Rapid 
Arithmetical project. He restricted himself to 
writing overviews of its architecture. Like the 
other projects at the institute, the Rapid 
Arithmetical Machine fell behind schedule. 

(U) Despite his patience and Caldwell's prom- 
ises, Deeds could not leave the future of his com- 
pany in the hands of an academic institution. 
Following Bush's suggestion, NCR established its 
own electronics research laboratory in the spring 

Page 26 

top gFPPFT</rnMiMT»pci m iiqa aiic, r am, naD Ann >m iiyi 


of 1938, 75 headed by Joseph Desch. Desch and his 
few assistants taught themselves about the latest 
electronic developments. 76 He completed an elec- 
tronic digital calculator by 1940 and explored the 
application of electronics to many types of busi- 
ness machines. 77 

Joseph Desch 

(Courtesy of 
the NCR 

Archive at the 

County, Ohio, 

(U) Then just as Desch's work was leading to 
the construction of hardware, the crisis in Europe 
and Deeds's patriotism ended Desch's commer- 
cial projects. His expertise in electronics and, as 
important, his unique manufacturing abilities, 
attracted the attention of the men in Weaver's 
group at the National Defense Research 
Committee. Before the end of 1940, Desch 
became part of the rise of Big Science. Within 
another year, he became central to the history of 
Bush's Comparator and to OP-20-G's future. 

(U) Notes 

1. (U) Karl L. Wildes and Nilo A. Iindgren, A 
Century of Electrical Engineering and Computer 
Science at MIT, 1882-1982 (Cambridge: MIT Press, 
1985), 7- Samuel C. Prescott, When MIT Was Boston 
Tech, 1861-1916 (Cambridge: Technology Press of 
MIT, 1954). 

2. (U) James M. Nyce and Paul Kahn (ed.), 
From Memex to Hypertext: Vannevar Bush and the 
Mind's Machine (Boston: Academic Press, 1991). 

3. (U) Larry Owens, "Straight Thinking: 
Vannevar Bush and the Culture ofAmerican 
Engineering," (Ph.D. Thesis, Princeton University, 
1987). Larry Owens, "Vannevar Bush and the 
Differential Analyser: The Text and Context of an 
Early Computer," Technology and Culture 27(1986): 
63-95. Montgomery B. Meigs, "Managing 
Uncertainty: Vannevar Bush, James B. Conant and 
the Development of the Atomic Bomb, 1940-45," 
(Ph.D. Thesis, University of Wisconsin, Madison, 
1982). Stanley Goldberg, "Inventing a Climate of 
Opinion: Vannevar Bush and the Decision to Build 
the Bomb," ISIS 83 (1992): 429. G. Pascal Zachary, 
"Vannevar Bush Backs the Bomb," Bulletin of the 
Atomic Scientists 48 (1992): 24. 

4. (U) James M. Nyce and Paul Kahn (ed.), From 
Memex to Hypertext: Vannevar Bush and the 
Mind's Machine (Boston: Academic Press, 1991), 
235-353, especially, Linda C.Smith, "Memex as an 
Image of Potentiality Revisited," 261-286. Adele 
Goldberg (ed.), A History of Personal Workstations 
(Reading, Mass.: ACM Press, 1988). On Bush's sci- 
ence policies in the post-WWII era, Daniel J. Kevles's 
preface to Vannevar Bush, Science: The Endless 
Frontier (Washington: NSF, circa 1992), ix. 

5. (U) Bush's patent history was traced through 
the historical files at the U.S. Patent Office's Crystal 
City, Virginia, facility. 

6. (U) Bernard Williams, "Computing With 
Electricity, i935-i945,"(Ph.D. Thesis, University of 
Kansas, 1984), 48. 

7. (U) Otto J. Scott, The Creative Ordeal: The 
Story of Raytheon (New York: Atheneum Press, 


8. (U) Vannevar Bush, Pieces of the Action. 
NDRC and its successor, the OSRD, were very elitist 
and Big Science oriented. James Phinney Baxter, 
Scientists Against Time (Cambridge, Mass.: MIT 
Press, 1968), and Irvin Stewart, Organizing 
Scientific Research for War: The Administrative 
History of the Office of Scientific Research and 
Development (Boston: Little-Brown, 1948). 

9. (U) Vannevar Bush, Pieces of the Action. 


Page 27 


10. (U) Larry Owens, "Straight Thinking: 
Vannevar Bush and the Culture of American 
Engineering," 289-90, 

11. (U) An overview of the results of his attempts, 
from the 1930s to the postwar era is found in MIT 
Archives, AC4 Boxes 30 and 36, Center of Analysis. 

12. (U) Vannevar Bush, "Instrumental Analysis," 
Bulletin of the American Mathematical Society, 42 
(October, 1936): 649. Karl L.Wildes and Nilo A. 
Lindgren, A Century of Electrical Engineering and 
Computer Science at MIT, 1882-1982, 230-3. 

13. (U) Smithsonian History of Computers 
Interviews, "Gordon S. Brown," January 27, 1970, 
provides a fascinating overview of many of the efforts 
at the Institute in the 1920s and 1930s. 

14. (U) There were several companies that made 
similar devices for gun control systems for the mili- 
tary, however. See the Barber-Coleman and 
Hannibal Ford companies. Note that Ford was inter- 
ested in building aversion of an analyser and, per- 
haps, donating it to Cornell University. See press 
releases by the Sperry Corporation, "Hannibal Ford," 
"Ford Instrument Company." On Ford and Bush, 
Rockefeller Archives, RG12.1 Diaries of Warren 
Weaver, March 3, 1935. 

15. (U) We have yet to have a technical history of 
the "tab" era that shows how they were used, but gen- 
eral overviews of needs and demand are found in 
James R. Beniger, The Control Revolution 
(Cambridge, Mass.: Harvard University Press, 1986), 
and in the brilliant Martin Campbell- Kelley, 
"Industrial Assurance and Large-scale Data 
Processing," Technohistory of Electrical 
Information Technology (Munchen: Deutsches 
Museum, 1991). 

16. (U) Irene S. Farkas-Conn, From Documen- 
tation to Information Science (New York: 
Greenwood Press, 1990). 

17. (U) Larry Owens, "Straight Thinking: 
Vannevar Bush and the Culture of American 
Engineering," 78. A valuable insight into the new 
Analyzer project is Charles Babbage Institute, 
Interview by William Aspray with Dr. Frank M. 
Verzuh, February 20 and 24, 1984- 

18. (U) Bush's acceptance of digital calculation, 
as evidenced by the plans for the electronic calcula- 

tor, the Selector, and the Comparator, calls into the 
question the thesis that he was wedded to analog 
models and calculation. See Larry Owens, "Vannevar 
Bush and the Differential Analyser: The Text and 
Context of an Early Computer," Technology and 
Culture 27(1986): 63-95. 

19. (U) NSA RAM File, Hooper toOP-20-G, 
"Cryptanalytic Machines," September 26, 1930, and 
Library of Congress, Papers of Stanford Caldwell 
Hooper, Box 18, Hooper to Secret Naval Board, "Staff 
Corps Personnel," February 7, 1936. 

20. (U) Library of Congress, Papers of Stanford 
Caldwell Hooper, "Memorandum on Johns Hopkins 
Group Visit," November 3, 1937, Box 17, 

21. (U) Rockefeller Archives RG12.1, Diaries of 
Warren Weaver, "Visit of Hooper and Dammon," 
June 10, 1938. 

22. (U) Wheeler, Yankee from the West, 18-20, 
386, and Farago, The Game of the Foxes, 477-8. 

23. (U) The ONR became a blessing to the uni- 
versities after World War II when it replaced the 
NDRC to subsidize research until the National 
Science Foundation was created. Harvey M. 
Sapolsky, Science and the Navy; The History of the 
Office of Naval Research (Princeton, NJ: Princeton 
University Press, 1990). The ONR is put into per- 
spective in Thomas A Guniston and Roger L. Geiger 
(ed.), Research and Higher Education: The United 
Kingdom and the United States (Buckingham: Open 
University Press, 1989), 3-17. 

24. (U) Paolo E.Coletta (ed.), The American 
Secretaries of the Navy, Vol. Ill, 1913-72 (Annapolis: 
Naval Institute Press, 1980), 663, dates the height of 
the conflict in 1933-34- But Harvey M. Sapolsky, 
"Academic Science and the Military: The Years Since 
World War II," in Nathan Reingold (ed.), The 
Sciences in the American Context (Washington: 
Smithsonian Institution Press, 1979), 379"399> 
describes a longer battle. 

25. (U) Harold G, Bowen, Ships, Machinery and 
Mossbacks: The Autobiography of a Naval Engineer 
(Princeton, NJ: Princeton University Press, 1954), 

26. (U) Ibid, 45. Bowen's role in the evolution of 
the ONR is traced in Harvey M. Sapolsky, Science 
and the Navy: The History of the Office of Naval 

Page 28 



Research (Princeton: Princeton University Press, 


27. (U) NARA RG457: SRH-150 "Birthday of the 
Naval Security Group"; SRH-305, "The Undeclared 
War: History ofRI," byLaurance Safford; SRMN- 
084 "Evolution of the Navy's Cryptologic 
Organization;" and SRH-152 "Historical Review of 
OP-20-G, 17 February 1944." 

28. (U) David Kahn, "Pearl Harbor and the 
Inadequacy of Cryptanalysis," Cryptologia 15(1991): 


29. (U) Louis Kruh, "Tales ofYardley: Some 

Sidelights to His Career," Cryptologia 13(1989): 
327-356. NARA RG457, SRMD-018 "Mexican 
Intercept Messages 1912-1924 MI-8." (U) NSA CCH 
Series XI K, Sam Snyder, draft of proposed history 
"Machines in U. S. Cryptology Before World War II," 
27 June 1975- (U) NSA CCH Series XII Z, 
"Memoranda on SIS, Formation of Cryptanalytic 
Group" from CCH Series XI K, Box 13, circa 1929- 


30. (U) More and more cases of important 
"acquisitions" during the 1920s and 1930s are com- 
ing to light. See Parker, Pearl Harbor Revisited: 
United States Navy Communications Intelligence, 
1924-1941 , 12. 

31. (U) Interviews and correspondence with 
Waldron S. MacDonald, 1987-91. MacDonald stated 
that Bush was the one that convinced the navy to 
investigate highspeed devices. It is more than likely 
that Bush was in touch with Hooper before 1930 
about such matters. See also Library of Congress, 
Papers of Stanford Caldwell Hooper, Box 21, August 
17, 1945, "Rough Draft of Comment," 3. 

32. (U) A useful history of the introduction of 
these machines is NARA RG457, SRH-004, "The 
Friedman Lectures on Cryptology." 

33. (U) Deavours and Kruh, Machine 
Cryptography and Modern Cryptanalysis , 212, 218. 
NARA RG457, "The Undeclared War: The History of 
RI," 15 November, 1943, by Laurance F. Stafford, 
Captain, U.S. Navy, and SRH-355, "Naval Security- 
Group History to World War II," 161. A History of 
Communications Intelligence in the United States 
With Emphasis on the United States Navy (NCVA), 
12. A wonderful insight in OP-20-G's methods is in 

Lt. L.F. Safford, "The Functions and Duties of the 
Cryptologic Section, Naval Communications," 
Cryptologia , 16 (1992): 265-281. 

34. (U) NSA RAM File, Hooper toOP-20-G, 
"Cryptanalytic Machines," September 26, 1930. 

35. (U) A book that overstates the case against 
the historical importance of formal analysis but 
which is still useful is Nigel West, The SIGINT 
Secrets: The Signals Intelligence War, 1900 to 
Today, Including the Persecution ofGordon 
Welchman (New York: 1986). Avery revealing and 
important document for the history of OP-20-G and 
American cryptanalysis is found in Louis Kruh, "Why 
Was Safford Pessimistic about Breaking the German 
Enigma Cipher Machine in 1942?" Cryptologia 14 
(1990): 253. 

36. (U) NARA RG457, SRH-355, "Naval Security 
Group History to World War II," 80. C. A. Deavours, 
"The Black Chamber: A Column: LaMethode Des 
Baton," Cryptologia 4(1980): 240-247. 

37. (U) Hooper's power to do this may have been 
based on the connections he established earlier in his 
career when he was the head of the Bureau of 
Engineering's new radio-sound division. 

38. (U) NARA RG457, SRH-355, "Naval Security 
Group History to World War II," 80. U.S. Navy- 
Office of Information, Biographies Branch, 13 
February 1958, "R. Adm. J. N. Wenger, USN, Ret." 

39. (U) NSA RAM File, OP-20-G to Chief of 
Naval Operations, "Cryptanalytic Machines - 
Photocells," November 11, 1931. 

40. (U) Photoelectric sensing for "sorting" has a 
long and complex history. See, for example, the 
patents of Michael Maul of Berlin dating from at least 
1927 which were assigned to IBM. See U.S. patents 
2000403-4. A Westinghouse engineer created an 
optical card sorter that caught Hooper's interest, 
Electronics 3 (October, 1931), 157. The 1930s work of 
the German, Emanuel Goldberg, who also invented 
the microdot, became of great significance to Bush's 
plans after World War II. Michael K. Buckland, 
"Emanuel Goldberg, Electronic Document Retrieval, 
and Vannevar Bush's Memex," JASIS 43 (1992): 284. 

41. (U) Among Hooper's and Wenger's recom- 
mendations was the exploration of the new statisti- 
cal-mathematical techniques being used in the 


Page 29 


advanced sciences. Although professional mathe- 
maticians were not brought into 0P-20-G until the 
onset of World War II (such as Howard Engstrom, 
Andy Gleason, and Marshall Hall), at least the 
younger men at OP-20-G were sent back to school for 
classes in statistics in the mid-iQ30s. NARA RG457, 
SRH-355, "Naval Security Group History to World 
War IT," 268. The influence of Lester Hill during the 
prewar years remains to be traced. 

42. (U) By the early 1930s, some scientists were 
using tabulating machines for advanced calculating. 
G.W. Baehne (ed.), Practical Applications of the 
Punched Carxi Methods in Colleges and Universities 
(New York: Columbia University Press, 1935) gives 
an insight to some of the uses and some of the special 
devices attached to the tabulators. Also useful for an 
understanding of precomputer calculation are 
William Aspray (ed.), Computing Before Computers 
(Ames, Iowa: Iowa State University Press, 1990), and 
Arthur Norberg, "High Technology Calculation in the 
Early 20th Century: Punched Card Machinery in 
Business and Government," Technology and Culture 
31 (1990): 753. 

43. (U) NSA RAM File, McClaran to Director of 
Naval Communications', January 7, 1932. NARA 
RG457, SRH-355, "Naval Security Group History to 
World War II," 75. The code was put into operation 
in December of 1930 and, luckily for the Americans, 
used until late 1938. 

44. (U) The pressures onOP-20-G multiplied 
because of a bizarre occurrence in 1930-31. The for- 
mer head of the State Department's and Signal 
Corps' cryptanalytic agency, Herbert Yardley, pub- 
lished his infamous book, The American Black revealed the United States' ability to 
read various Japanese code and, perhaps, cipher sys- 
tems. NARA RG457» SRH-151, "Military Study: 
Communication Intelligence Research Activities," 9. 

45- (U) NSA RAM File, "McClaran to Director of 
Naval Communications," January 7, 1932. 

46. (U) NSA RAM File, Huckins to Bureau of 
Engineering, "IBM Rental," May 15, 1933. 

47. (U) Deavours and Kruh, Machine. 
Cryptography and Modern Cry ptanaly sis , 212. 
Hooper was especially worried about Britain's new 
shipboard cipher machines in the early 1930s. NSA 

RAM File, Hooper to OP-20-G, "Cryptanalytic 
Machines," September 26, 1930. 

48. (U) NARA RG457, SRMN-084, "The 
Evolution of the Navy's Cryptologic Organization," 
and SRH-264, "A Lecture on Communications 
Intelligence," by Capt. J. N. Wenger, USN, August 14, 

49. (U) NARA RG457, SRH-151, "Military Study: 
Communication Intelligence Research Activities," 

50. (U) NARA RG457, SRMN-083, "Military 
Study of Secret Radio Calls, January 1938." 

51. (U) Howeth, History of Communications 
Electronics in the United States Navy; With an 
Introduction by Chester W. Nimitz , 538. 

52. (U) John C. Walter, "William Harrison 
Standley," in Robert William Love, Jr. (ed.), The 
Chiefs ofNaval Operations (Annapolis: Naval 
Institute Press, 1980), 93. 

53- (U) NARA RG457, SRH-355, "Naval Security 
Group History to World War 11," 99. 

54. (U) Vannevar Bush, Pieces of the Action (New 
York: Morrow, 1970), 71, and Library* of Congress, 
Papers of Stanford Caldwell Hooper, Box 16, 
"Binaural Sons of the C," June 1,1934. One of 
Hooper's most valuable connections with the scien- 
tific elite was the "alumni" club for those who had 
worked at the two major sonar development sites 
during World War I. He quite possibly met with Bush 
in its informal context. He certainly had later con- 
tacts with Bush when they were both associated with 
the NACA. MIT was a major training resource of the 
navy during WWI, and it had the Pratt School of 
Naval Architecture. Karl L. Wildes and Nilo A. 
Lindgren, A Century of Electrical Engineering and 
Computer Science at MIT, 1882-198 2 (Cambridge: 
MIT Press, 1985), 393. 

55. (U) NSA was unable to provide copies of the 
four Bush reports and the rest of files on Bush's work 
for the navy in the 1930s. As will be discussed, the 
later four general reports were for the design of the 
machine that became known as the Comparator. 
Bush's initial report to Hooper and Wenger was 
probably much more general and was most likely 
concerned with very broad issues of communications 
technology. Bush's oral history version of the negoti- 




ations does not quite fit with other evidence, MIT 
Archives MC143 ma to 116. 

56. (U) A useful long-term view of academic-mil- 
itary relations is Henry Etzkowitz, "The Making of An 
Entrepreneurial University: The Traffic Among MIT, 
Industry, and the Military, 1860-1960," E. 
Mendelsohn, etal., (ed.), Science, Technology, and 
the Military 12 (1988): 524. 

57. (U) Ronald W. Clark, The Man Who Broke 
Purple: The Life ofthe World's Greatest 
Cryptologist, Colonel William F. Friedman (London: 
Weidenfeld and Nicolson, 1977). 

58 . (C) David P. Mowry, "Listening to the Rum 
Runners," Cryptologic Quarterly, 2 (1983), 27-50. 

59. ity- Friedman seems to have found away to 
make sure that Yardley would be unable to join the 
new army group. NSA CCH Series XII Z, 
"Memorandum on SIS, Formation of Cryptanalytic 
Group" from CCH Series XI K, Box 13, circa 1929- 
1939. David P. Mowry, "Listening to the Rum 

60. (U) NSA CCH Series XI K, Sam Snyder, draft 
of proposed history "Machines in U.S. Cryptology 
Before World War II," 27 June 1975. 

61. (U) NSA CCH Series XII Z, William F. 
Friedman, "Addenda on the IBM Sorter," circa 
August 1935- 

62. -(€)■ NSA CCH Series XII Z, William F. 
Friedman, "Invention of a Cryptanalytic Coincidence 
Counter," Signals Intelligence Section, 14 April 1937. 
NSA CCH Series XII Z, William F.Friedman, 
"Description ofthe General Principles of an 
Invention for Locating Idiomorphs and Isomorphs in 
Cryptanalysis," 14 April 1937. 

63. (U) Robert F. Kohler, Partners in Science: 
Foundations and the Natural Sciences, 1900-1945 
(Chicago: University of Chicago Press, 1991). Robert 
F. Kohler, "The Ph.D. Machine: Building on the 
Collegiate Base," ISIS 81 (1990): 638-662. 

64. (U) Robert F. Kohler, Foundations and the 
Natural Scientists, 1900-1945. Roger L. Geiger, To 
Advance Knowledge: The Growth of American 
Research Universities, 1900-1940 (New York: 
Oxford University Press, 1986). 

65. (U) Larry Owens, "Straight Thinking," 78-79. 

66. (U) Rockefeller Archives, RG12.1, Diaries of 
Warren Weaver: May 1, 1940 "Atanasoff Visit"; 
October 24, 1939, "Tour of Computing Centers"; 
"Visit to Boston," October 29, 1939; October 5, 1939, 
"Howard Aiken Visit"; May 24, 1939, "Visits of 
Harrison and Caldwell"; and January 1, 1939, "Visit 
to MIT." 

67. (U) Daniel J.Kevles, The Physicists: The 
History ofthe Scientific Community in Modern 
America (New York: Knopf, 1978), 296. 

68. (U) An important article on the history of 
both mechanical and electric-electronic fire control 
devices is A. Ben Clymer's, "The Mechanical Analog 
Computers of Hannibal Ford and William Newell," 
Annals ofthe History of Computing 15 (1993): 19-34. 

69. (U) Carroll Pursell, "Science Agencies in 
World War II: The OSRD and Its Challengers," in 
Nathan Reingold (ed.), The Sciences in the American 
Context: New Perspectives , 359. 

70. (U) On Deeds, Isaac F. Marcossen, Wherever 
Men Trade (New York: Dodd-Meade, 1948), and 
Colonel Deeds: Industrial Builder (New York: 1947). 

71. (U) Copies of his patents were located in 
Bush's files at the Library of Congress. 

72. (U) Library of Congress, Papers ofVannevar 
Bush. Bush had copies of various Hofgaard patents in 
his papers. 

73- (U) There is no evidence on whether or not 
Bush and Deeds knew that IBM was beginning to 
explore electronic calculation, but they must have 
been aware, because of patent claims, of IBM's grow- 
ing interest in microfilm and allied devices, including 
a "statistical" machine. Hagley Museum and Library, 
Accession 1825, Honeywell vSperry-Rand Trial 
Records, Chronological File, March 1937, letters re: 
visit of Green and Sullivan to MIT to view electronic 

74. (U) Hagley Museum and Library-, Accession 
1825, Honeywell v Sperry-Rand Trial Records, May 
19, 1938, Bush to Deeds "Center of Analysis." A gen- 
eral overview of the machine and project is in Brian 
Randell (ed.), The Origins of Digital Computers: 
Selected Papers, 3rd ed., 294, and Bernard Williams, 
"Computing With Electricity, 1935-1945," (Ph.D. 
Thesis, University of Kansas, 1984), 137-170. 


Page 31 


75. (U) Hagley Museum and Library', Accession 
Records, Deposition of Joseph Desch; Electronics 
Research that Bush asked if he wanted a job at NCR. 
Honeywell v Sperry Rand Trial Records "Report of 
Joseph Desch on Electronics Laboratory to H.N. 
Williams." August 16, 1938, Eugene Kniess, "First 
Lab Rediscovered," NCR Dayton 6(1973): t-3. 

76. (U) Hagley Museum and Library, Accession 
1825, Honeywell vSperry-Rand Trial Records, 
Desch Deposition, "Report of Joseph Desch on 
Electronics Laboratory to H. N. Williams," August 16, 

77. (U) Hagley Museum and Library, Accession 
1825, Honeywell vSperry-Rand Trial Records, 
Desch Deposition, and Reports of April 28, 1939 and 
March 25, 1940. Smithsonian Interviews with Desch 
and Mumma. 


top ocorcrn/oo MiN TOmiL to uoa, auc, ca n , g b r amp n zu;xi 

Chapter 2 
(U) The First Electronic Corputer: Perhaps 

(U) A Reminder of Hooper's Hopes and 

(U) The development of new cipher machines 
and the maturation of radio led to a critical data 
problem for America's cryptanalysts. There was 
more and more data, and it was overwhelming 
those who were charged with turning it into use- 
ful information for policymakers. The failure to 
predict the attack on Pearl Harbor, for example, 
was the result of too much data. The thousands of 
intercepted Japanese naval messages could not 
be analyzed with the men and equipment avail- 
able to Laurance Safford's OP-20-G. 1 

(U) Vannevar Bush realized the similarity 
between the challenges facing the cryptanalysts 
and the ones faced by those who were trying to 
reform the way the nation handled scientific 
information. He believed the two groups could 
share technology and methods. Captain Stanford 
C. Hooper might not have been aware of the 
trends in scientific literature, but he was certain- 
ly frightened by increasingly sophisticated cipher 
machines being introduced by potential enemies. 
That was what led him and his protege, Joseph 
Wenger, to Bush in late 1935. 2 Despite Hooper's 
vision and Wenger's efforts, OP-20-G began 
World War II without any operating high-speed 
devices. The Rapid Analytical Machine project 
had to begin over again in 1942 and in conditions 
ill-suited to long-term development. 

(U) The reasons for the failure of Hooper's 
1930s plans for the application of 
scientific/mathematical methods tocodebreak- 
ing are complex. Bureaucratic tangles, bad luck, 
personality clashes, Bush's stubbornness, inter- 
national crises, and the intransigence of technol- 
ogy partially account for the lost opportunity. But 
the major factor was institutional. Above all else, 

the military had not yet placed great faith in the 
kind of information that cryptanalysis or other 
signals intelligence could provide. 3 

(U) The Institutional Context 

(U) By the mid-1980s, Hooper and his admir- 
ing young officers feared that America would be 
dragged into a war while Naval Communications 
was unprepared for a face-off with any power. 
Hooper's 1930s strategy, to collaborate with uni- 
versities and corporate centers, was an attempt to 
compensate for the lack of money needed to pre- 
pare for a modern war. The Chief of Naval 
Operations supported his plans, but the CNO 
approval did not mean smooth sailing for Hooper 
and his men. To Hooper's regret, OP-20-G con- 
tinued to have to depend on the Bureau of 
Engineering because navy law and "G's" pauper 
budgets allowed little else. More independence 
and money might have come to OP-20-G if there 
had been widespread faith in signals intelligence. 
But despite the contributions of Herbert Yardley's 
Black Chamber during the 1920s, then OP-20-G's 
penetration of Japanese naval codes, and then the 
cracking of Japan's diplomatic messages, code- 
breaking remained a stepchild of the American 
military. 4 Ironically, the reading of the Japanese 
naval and diplomatic code and cipher systems 
during the 1920s and 1930s masked the need for 
the long-term programs required for the develop- 
ment of advanced methods and machines. Even 
the navy's operating cryptanalysts did not lobby 
for such a program. 5 Only two men, Hooper and 
Wenger, saw the need and were willing to suffer 
the possible career penalties imposed on those 
who became advocates for unpopular causes. 

(U) Hooper and Wenger had never aban- 
doned their 1930 hopes for machines that would 


Page 33 


be much more advanced than the tabulators. 6 In 
late January 1936, Wenger met with Bush and 
discussed OP-20-G's hopes and problems. 7 Bush 
presented Wenger with a handwritten eight-page 
outline of his plan for automating OP-20-G's 
cryptanalytic section. 8 Within a week, Wenger 
had secured the new Director of Naval 
Communications's approval of the proposed rela- 
tionship with Bush. 

(U) TJie First Defeat; Bush Is Rejected 

(U) Just as Wenger proudly submitted his 
own visionary outline for the reorganization of 
OP-20-G, he received a slap in the face. The 
Bureau of Engineering refused to approve the 
agreement with Vannevar Bush! 9 There was rea- 
son for the bureau's alienation. What Bush 
demanded and what Hooper and Wenger agreed 
to were startling. Bush demanded having the gov- 
ernment pay the bill while he remained free of 
supervision. He wanted the relationship with the 
navy to match the ideal relationship between uni- 
versity researchers and major private founda- 
tions. The researcher would submit a general pro- 
posal and then be funded without any interfer- 
ence from the grantor. Following on his beliefs, 
Bush had refused to sign a typical navy contract 
or to make any promises about the results of his 

(U) In addition, the original understanding 
did not include a promise to construct any 
machinery. Bush and Wenger had also agreed to 
ignore the regulations demanding competitive 
bidding on naval contracts. In addition, Bush 
requested what was an enormous amount of 
money in the era, at least for the navy. To hire 
Bush meant taking precious resources from the 
bureau and from OP-20-G. 

(U) A Machine Too Soon 

(U) There were also serious technical objec- 
tions. Although only the barest sketch, Bush's 
early 1936 proposal showed that he wanted the 

navy to use optical scanning, high-speed data 
tapes, electronic computing, and microfilm in a 
series of increasingly complex cryptanalytic 
machines. Such technologies, Bush emphasized, 
would allow processing speeds from ten to one 
hundred times faster than the tabulators. 
Engineering thought that his recommendations 
were speculative and liable to be very costly fail- 
ures. Engineering's staff had good reason to be 
worried about the technical ideas. The core tech- 
nologies Bush recommended were, to significant 
degrees, still experimental. 

(U) Also, the bureau's engineers claimed they 
had their own solution to the problem of auto- 
matic cipher machines. They were reluctant to 
give Wenger even a hint of their approach, how- 
ever. 10 Whatever its secret alternative to Bush's 
proposals, engineering had accepted the tabula- 
tor. It was an off-the-shelf technology that had a 
stable manufacturer. IBM knew the ropes of gov- 
ernment contracting and was investing in ongo- 
ing development with its own funds. Many of 
engineering's men were already creating signifi- 
cant and clever modifications to IBM's machines, 
making them more effective cryptanalytic tools. 

(U) In addition, the views of OP-20-G's crypt- 
analysts were not incomplete harmony with 
Wenger's. The operational cryptanalysts won- 
dered who could steal the time to devise the new 
procedures necessary to make such strange tech- 
nology useful. By the mid-i930s, Laurance 
Safford and Jack Holtwick became more allies 
than enemies of Hooper's long-term plans, but 
the remainder of the staff were willing to join with 
engineering in seriously questioning the value of 
Bush's machines." All the objections and emo- 
tions meant that by mid-1936 the attempt to 
bring electronics to American cryptanalysis was 
deadlocked, if not defeated. But Stanford Hooper, 
Vannevar Bush, and Joseph Wenger collected the 
needed political support, drew up a new plan, and 
outflanked the bureau and the conservative crypt- 

Page 34 


IUP ShCKb l /ilUMINT/m E L TO U3A, AU3, CA N , ODR A N D H2U/X1 

(U) Hooper and Wenger developed anew 
strategy to surmount any remaining objections. 
To placate the engineers, Hooper agreed to ask 
Bush to submit a more detailed and specific pro- 
posal. The new Bush proposal was submitted to a 
special research group in the navy rather than to 
engineering. In September 1936, within a week 
after he received the new plan, Hooper reported 
to Bush that the prestigious research board had 
approved his project. Wenger and Bush devel- 
oped compromise positions on the bureaucratic 
and legal objections, then presented the new pro- 
posal to engineering. The bureau gave in, but it 
took almost all of October and November 1936 to 
draft an acceptable contract. 

(U) Under 1937's formal contract, Bush 
agreed to focus on the details of a particular 
device so that engineering could have something 
concrete. He was to submit four reports, each 
detailing a major component of the proposed 
machine. The commitment to details and the year 
and one-half time limit for delivery of all the 
reports helped to satisfy the bureau's demand for 
a scheduled product. 12 

(U) The Decision to Build a Machine 

(U) Bush had become attached to Wenger and 
Hooper, and their pleas convinced him to make a 
gentleman's promise that he soon regretted. He 
told them he would try to build a machine, and if 
he succeeded he would give it to the navy at no 
additional cost, except for shipping charges for 
the finished machine. 13 It had become very 
important to Wenger to have a device. To ensure 
that his project would not die when Bush's con- 
tract ended, Wenger needed a machine to prove 
that photoelectronics was practical. 

(U) Bush was not sure that he could build a 
machine in time, but in early 1937 he was 
absolutely sure of one thing: MFTs work for OP- 
20-G would be cut off by mid- 1938 when the con- 
tract with the bureau terminated. During the year 
of bickering with the navy, Bush became involved 

in an increasing number of projects that were 
critical to the Institute's planned analysis center 
and his career. One consequence was that the 
navy's project became more of a burden than an 

(U) Bush spent much time on the initial 
designs for an astounding general-purpose elec- 
tronic digital computer. He sent his students and 
colleagues the first of several outlines of the pro- 
posed digital device, soon to be called the Rapid 
Arithmetical Machine, in January 1937. 14 In the 
three years after the first contacts with the navy, 
Bush and his men had put all the years of struggle 
behind them. Bush had his "boys" immersed in 
three highly innovative digital projects: the elec- 
tronic Rockefeller Analyser; the electronic, pro- 
grammable Rapid Arithmetic Machine; and the 
Rapid Selector. 

(U) Bush and Wenger Select a Problem 

(U) Bush consulted with Joseph Wenger and 
opted for a device to help OP-20-G apply the lat- 
est statistical techniques to the cipher problems. 15 
Bush knew that if a machine was built, it had to be 
one that was reliable enough to convince the 
bureau to fund a long-term RAM project. 
Furthermore, Bush knew that any machine he 
created would have to outperform OP-20-G's tab- 
ulators and the special mechanical devices l6 that 
had become so dear to many of its staff. His 
machine had to be much faster than the electro- 
mechanical devices. 17 

(U) There were many advanced cryptanalytic 
methods for Bush to select from. Perhaps 
unknown to Bush or Wenger, the United States 
Army's cryptanalyst, William F. Friedman, was 
toying with ideas about the use of optical scan- 
ning. In April 1937, just as Bush was filling in the 
design of his machine, Friedman filed a patent for 
a system. The application did not mention crypt- 
analysis, and its examples of possible use were 
related to analog business applications, such as 
the sorting of packages, but Friedman must have 

i a P SECR ET /je O M I N T /mEL TO UflA, AU3, OA N , ODR A N D N ZU/X1 

Page 35 


realized that optical scanning had great potential 
for cryptology. 18 Despite such projects, Bush was 
facing the great challenge of creating what was 
the world's first high-speed cryptanalytic 
machine. Balancing all the factors, including his 
almost unshakable commitment to the three 
technologies of film, optics, and electronic count- 
ing, Bush decided to automate one of the most 
central new statistical methods, the Index of 

(U) The Index 

(U) The method Bush and Wenger selected 
for the machine, the Index of Coincidence, was 
the most ubiquitous of the new theoretically jus- 
tified statistical procedures. It was a formal and 
universal method that could not be made worth- 
less by a slight change in a cipher system. It was 
based on the laws of probability. The Index was 
rugged and independent because it needed only 
intercepted cipher text and because it could 
attack any type of cipher system. 19 It also had a 
wide range of powers. 

(U) The Index allowed an analyst to identify 
messages or portions of messages that were pro- 
duced by the same settings of an encryption 
device. That was a first step to determining the 
wiring and settings of the encrypting components 
of the machines. The Index of Coincidence could 
then be put to work to identify a cipher key or the 
order of the cipher wheels in a machine. Such new 
methods were essential to an independent attack 
on the cipher devices. The stepping switch and 
wired-wheel machines, such as the Japanese 
Purple and the German Enigma, were designed to 
be unbeatable. They had cascades of transposing 
rotors which repeatedly changed one letter to 
another. Although each rotor was simple, togeth- 
er they produced a long sequence of letter substi- 
tutions without repetition or pattern. 

(U) Such machines as Red, Purple, and the 
Enigma came close to creating a random 
sequence, but not quite. They appeared to be ran- 

dom because of the length of the cycle of unique 
substitutions created by the three or four rotating 
enciphering wheels or switches. But after 26 x 26 
x26 or more rotations, the wheels returned to 
their initial positions, and the machine began to 
repeat its letter substitutions. That made them 
technically nonrandom and allowed many 
nations to use Index methods against the simple 
Enigmas ofthe Spanish Civil War. 20 However, 
every nation was improving its cipher machines. 
Additional wheels with unique transpositions, 
varied latches that turned a neighboring wheel 
erratically, and plugboards to further disguise a 
machine's input-output relationships were added 
to many devices. The combinations of wheels, 
wheel settings, and plugboard links meant that 
trillions of possibilities had to be explored. 

(U) In response, cryptanalysts countered with 
various forms of automation. But most, like 
Poland, bet on limited methods and machines, 
ones to exploit the quirks of particular cipher 
machines or the procedural errors ofthe enemy. 
There was good reason for such a turn away from 
science. The German specialists in charge ofthe 
Enigma, who were aware ofthe laws of probabil- 
ity and also ofthe speed of film and optical 
machines, were confident that it would take any 
formal attack too long to be of use to an enemy. 
Given the special defenses built into the Enigma, 
they calculated that it would take any machine so 
long to perform a statistical analysis that by the 
time a setting was identified, its messages would 
be of no military value. 21 

(U) When Wenger met with Vannevar Bush in 
1937 to decide exactly what type of machine to 
design, his goal was the creation of a device so 
rapid that pure statistical analysis would be prac- 
tical. After balancing the needs of OP-20-G and 
the technological possibilities, he and Bush decid- 
ed to automate the heart ofthe IC method, coin- 
cidence counting. 

(U) A coincidence was the appearance ofthe 
same letter in the same relative position in two or 

Page 36 


■mo ccrpc-n/^MmTi/pci m iiqa aiiq cam rtRP Ann M7I »Y1 

more messages or in an offset of two copies of the 
same message. The method could be extended to 
the identification and counting of more than sin- 
gle letter matches, but the essence of the Index 
was the counting of single coincidences. If the 
number of matches exceeded the number expect- 
ed from a random distribution of letters, then 
both messages were probably a product of the 
same wheels, wheel settings, and portion of the 
encryption machine's cycle. 

(U) As the enciphering machines became 
more complex, the Index developed an almost 
insatiable demand for data. The Index could be 
computed with electromechanical machines, 
such as a counting sorter or a tabulator with addi- 
tional relay circuits. But even with the IBM 
machines, the process was very slow and labor 
intensive; a long message could take days to ana- 
lyze. One of the reasons the Index was selected as 
the method for Bush to automate was that it was 
so difficult to perform on electromechanical 

(U) An Added Bonus, Possibly 

(S// S I//REL) Wenger and Bush were com- 
mitted to mechanizing the IC method, and both 
wanted to encourage the navy's codebreakers to 
apply mathematics, but Wenger realized that the 
operating codebreakers had to use some less than 
"scientific" approaches. If Bush could automate 
them, the MIT machine and statistical methods 
might receive a friendly evaluation by the crew at 
OP-20-G. Bush agreed to sketch machines for 
those rather crude methods, and he hinted that 
he would try to have the proposed Comparator 
(for the IC) be able to perform two of them. Both 
methods, the Brute-force search and 
Symmetrical-sequences, asked for a search 
through massive amounts of data to "locate," not 
count, coincidences. The coincidences sought 
were not based on individual letters, but matches 
between relatively long strings of cipher text or 
long strings of text whose letters had been trans- 
posed into their position relative to the starting 

letter of the string. Both approaches were ways to 
identify messages that were likely to have been 
produced by the same key. They were used to find 
messages that were in "depth." No mathematics 
was required; a machine just had to sense the 
long coincidence and then inform its operator 
where it was located. 

(U) Bush Outlines the Machine and Sets 
Difficult Goals 

(U) After the navy contract was signed in 
January 1937, Bush took time away from his 
other duties to work on the architecture of his 
Index machine, the Comparator. He decided to 
divide the project into four major parts corre- 
sponding to functional units of the proposed 
machine. Then, he chose what hardware was to 
be used in each. Last came an equally challenging 
step, finding the four men he needed to fill out his 
sketches and, perhaps, build a machine. 

(U) Bush had a frustrating time finding quali- 
fied men. The need for secrecy made it almost 
impossible to locate men and still maintain good 
relations with the faculty. Only three people at 
MIT, really two, knew what the work was for. 
Bush and the project manager knew details, but 
MITs president learned only that secret work was 
in progress. The men who were to build the com- 
ponents and their regular faculty supervisors 
were not told of the navy connection. Once 
employed, they were instructed to be confidential 
about their work but not told why. They would 
never be informed as to what their components 
were for. 22 

(U) Two graduate students received the initial 
assignments. Jerry Jaeger, who had aback- 
ground in machine tools and automatic controls, 
was given the first task, to build the critical input 
mechanism. Richard Taylor, who was already 
important to the Rockefeller project's electronics 
and who would soon take charge of the Center of 
Analysis, was chosen to be responsible for the 
electronic circuits. The third man, who was asked 


Page 37 

TOP 0C0Rnr//0O M I N T//nCL TO UOA, AUG, CAM, ODR and HZU/X1 

to develop the component to read the data tapes, 
was in a somewhat different position at the 
Institute than Jaeger or Taylor. Herbert E. Grier 
was a graduate of 1933 who remained at the 
Institute as an unpaid research associate. Bush 
was unable to find the needed fourth man among 
the student body. He turned to one of the 
Institute's machinists, Walter Kershner, to design 
and construct what seemed to be the least chal- 
lenging part of the Comparator, its data input 
device. Kershner probably had been working on a 
similar automatic tape punch for the Rockefeller 

(U) Finding a manager for the project was a 
greater challenge. It was not until early summer 
1937 that Bush thought he had a lead on a quali- 
fied engineer: Waldron Shapleigh MacDonald. 

(U) MacDonald was one of the most unusual 
and fascinating ofMITs students, and he 
remains an unrecognized figure in the birth of the 
modern computer. MacDonald first appeared at 
MIT in the early 1930s when he enrolled as a spe- 
cial undergraduate student. His initial year in 
Cambridge was spent trying to prove to the elec- 
trical engineering faculty that his lack of formal 
preparation was not a barrier to academic suc- 
cess. Although he performed well in his classes, 
he was unable to surmount bureaucratic hurdles, 
illness, and the depletion of his savings. He had to 
leave MIT without a degree. But he quickly found 
very well-paying work as an engineer and began a 
lifelong career as an innovator in computers and 
automatic controls. 

(U) Bush offered MacDonald a professional 
salary and help in obtaining a master's degree in 
electrical communications at the Institute. In 
return, MacDonald was asked for a firm commit- 
ment to come to MIT to see the navy's project 
through to completion. But MacDonald needed 
time to fulfill his existing responsibilities, and he 
did not arrive at MIT until September 1937, leav- 
ing only some ten months to become oriented, to 
check and revise the Comparator's parts, prepare 

reports, and assemble and test the historic 
machine. 23 

(U) MacDonald's ingenuity and his hands-on 
engineering ability were needed on the navy's 
1930s project, but his role was not a truly creative 
one. Well before he arrived in late 1937, the 
design of the machine and the schedule for the 
project had been determined. His job was to 
make what Bush had specified come to life and to 
do it before the end of the navy contract. 
Unfortunately for MacDonald, he inherited a 
fixed design, components which were hastily 
made by others, Bush's order to "get the job done 
on time," and full responsibility. By September 
1937 Bush was already too busy with his other 
work to attend to the now rather inconsequential 
navy project. Among other things, Bush was 
readying himself to assume the leadership of the 
powerful Carnegie Institution. 

(U) The Comparator Really Doesn't Go to 

(U//FQUQ) Bush and Wenger were very wise 
insetting the limited goal of a machine for the 
Index of Coincidence. Electronic computation 
was having its birth pangs, and no one had a way 
to create a machine whose hardware could be 
made to imitate any process. A major reason why 
all the 1930s computers were limited in function 
was the absence of a viable memory technology. 24 
A universal data computer, one that worked on 
large volumes of input and that had high-speed 
memory, did not appear until the 1950s. Then, 
machines such as the UNIVAC depended upon 
very demanding, slow, and expensive, magnetic 
tape memory systems 


(U//FQUO) Bush's first sketches of his 
Comparator reflected the limitations of the mem- 
ory and electronic technologies. Each of the 
Comparator's four major components had its own 
very significant practical challenges. The state of 
the technology did not allow elegant solutions to 
the problems of high-speed input, sensing, count- 

Page 38 

TOP SFf.RFT)/cnuiMT»PFi TniiQA fliiQ cam nao aim mti fnri 


ing, and recording. Because of the conduct of the 
1930's Comparator project and the nature of OP- 
20-G's early wartime efforts, it was not until late 
1943 that America had more than the patched-up 
Bush Comparator to represent its nearly fifteen 
years of attempts to build sophisticated electron- 
ic codebreaking devices. 

(U) Too Much to Ask of Mere Machines 

(T0//01//REL ) The Index was a demanding 
cryptanalytic method. To tally all the possible sin- 
gle letter coincidences in two messages calls for 
(n*(n-i)) comparisons. 26 If two four-letter mes- 
sages are examined for coincidences, twelve com- 
parisons must be made; 500 messages demanded 
almost 250,000 tests; a 2,000-letter message 
called for almost 4,000,000. Complete analyses 
of long messages could take days or weeks by 
hand and tabulator methods. Compounding the 
challenge of raw speed was Wenger's demand 
that the Comparator be able to handle the longest 
messages. There was good reason for that 
because the more characters in a message the 
more likely that something of value would emerge 
from an analysis. Fortunately, cryptologists 
around the world knew that messages with too 
many words posed a danger to their systems and 
instructed that messages be limited to as few 
words possible. The very upper limit was 2,000 
characters. Messages of 200 characters were typ- 
ical, but the need to analyze longer ones in a time- 
ly way made speed and a large memory important 
goals. 27 

(U) Combined with Bush's desire for a mini- 
mal number of electronic components, the call for 
speed created unexpected challenges for the stu- 
dents at MIT. One of them was printing. To main- 
tain speed, printing had to be done while the tape 
was running. The solution Bush and his men 
devised was sensible but crude, and it led to a 
need for an even faster mechanical tape drive. 
Printing was to take place while a blank portion of 
tape was running. Tn practice, this meant that 
approximately one half of each tape was blank, 

thus halving the number of possible comparisons 
during a am of the tape. Because of that, Bush's 
men had to double the originally planned speed 
of the drive to achieve the processing goals. 28 

(U) Even without the tape handicap, Bush 
had to outdo much existing technology to achieve 
his minimum Comparator speed. 29 Bush wanted 
the machine to deliver data to the reading station 
at over thirty times the rate of standard telegraph 
equipment and sixty times faster than a movie 
projector if it was to reach the goal of 20,000 
comparisons a minute. Even in the late 1940s, the 
most sophisticated high-speed transmission 
"baud rates" were in the range of 1,800 characters 
a minute - or more than ten times slower than 
Bush needed in order to make the navy machine 
an attractive alternative. There were special high- 
speed drives for sending bulk messages, and dur- 
ing World War II "flash" systems were developed. 
Those devices, however, were not proven in the 
mid-i930s. The talking picture industry did not 
provide much help. In the 1930s, moving picture 
film was moved at less than 300 feet per hour. 30 
The Comparator had to sense and route data at 
rates forty times greater than an IBM sorter and 
160 times faster than a tabulator. 

(U) Wenger thought that he might overcome 
the bureau's protests if Bush could add parallel 
features to his essentially serial machine. Wenger 
asked him to try to include what would be needed 
to make isomorphic and three- and four-letter 
(polymorphic) coincidence tests that had been 
discussed earlier. 

(T 8 //SI//REL) Wenger also gave his 
approval for the "locating* feature. It would allow 
what the World War II cryptanalysts called "brute 
force" searching. Masses of data could be scanned 
at every position of two messages with the hope of 
finding indications that two messages had been 
enciphered with the same key. 


Page 39 


(U) No Thanks for the Memories 

(U) Because the Comparator was a data- 
dependent machine, the greatest problem facing 
Bush's students was how to store and retrieve 
information. The Comparator needed a large- 
scale and very high-speed memory, but such 
memories did not exist in the 1930s. 31 What was 
on the technological horizon was not encourag- 
ing. Storage in massive banks of capacitors or 
resistors, which some computer designers were 
thinking of using, was too expensive, and such 
banks took too long to load and unload. 32 The 
rumors about the use of special versions of televi- 
sion tubes as memory were just that in the mid- 
19305. And no one thought that delay lines would 
ever be able to hold more than a few bits of infor- 
mation. In 1937 work was just beginning on mag- 
netic memories, and storage of large amounts of 
data in two or multistate electronic tubes or relays 
was out of the question. 33 

(U) Unfortunately for Bush and Wenger, 
there had been few advances in tape technology 
since the introduction of modern automatic tele- 
graph readers in the early twentieth century. 
Standard teletype technology had not evolved 
into a competitor to the punch card. 34 In early 
1937 the only option seemed to be microfilm. 

(U) Bush thought his men would overcome 
the difficulties caused by film shrinkage and dis- 
tortion when the film was sped past a reading sta- 
tion. 35 Unfortunately, microfilm proved too diffi- 
cult for a machine that could meet the mid-1938 
deadline for the delivery of the Comparator. As a 
result, in mid-1937 Bush sent his students on a 
hurried search for another medium and a way to 
move it at incredible speeds. The MTT men chose 
a unique 70mm-wide paper tape that Eastman- 
Kodak used for packaging its movie film. It was 
strong, wide enough to accommodate Bush's cod- 
ing scheme, and, very important, it blocked light 
because of its acetate coating and its alternate 
red-black layers. 36 Also, early tests indicated the 
tape would maintain its structural integrity after 

being punched. All those features justified the 
high cost of the Eastman product although it was 
soon learned that its data capacity would not be 
much more than that of telegraph tape. 37 

(U) The disappointingly low density meant 
that much effort had to be put into the develop- 
ment of a high-speed tape drive, one burdened 
with some very special demands. In addition to 
the need for ultra-high speeds, the tape transport 
had to pass two tapes in perfect alignment over 
the reading station, then step one tape one char- 
acter relative to the other until all possible com- 
parisons had been run. 38 

(U) The Limits ofMecJianics 

(U) The first man on the summer crew was 
given the responsibility of creating the mechani- 
cal combination needed to compensate for the 
low data-carrying power of the Eastman tape. 
Already familiar with the drives in the machines 
used in the cloth and newspaper industries, the 
young engineer decided to center his component 
on a four-foot long frame to hold the tapes. 
Pulleys were to maintain the required tension on 
the loops of tape. Driven by a fast electric motor 
and a system of shafts and gears, the tape was 
guided by both rollers and sprockets. 39 The entire 
transport was mounted on tall legs and stood 
some four feet off the ground to ease the chore of 
changing tapes. 40 

(U) The tape transport was well designed and 
was delivered on schedule, but it did not reach the 
speeds Bush desired. At its best moments it ran at 
less than two and a half miles an hour, not the five 
or more needed for a truly rapid machine. The 
tape was the machine's timer and set many of the 
requirements for the other major components. 
Once its features were known, work on the read- 
ing station and electronic counters could be com- 
pleted. Armed with Bush's previous instructions 
and the specifications for the tape drive, the next 
man tackled the problems of photoelectric 

Page 40 



(U) Let There Be Light, But Not Too Much 

(U) One of Bush's first technical commit- 
ments was to the sensing of the presence of light 
rather than its absence. Following on that, he 
ordered his men to code each letter of a message 
by punching a hole in a column of the seventy- 
millimeter wide tape. There was to be only one 
hole to a column of twenty-six fields. An addi- 
tional field in each column served as a timer. If a 
column held data, this extra field was punched. 
When two active columns overlapped, light was 
directed to a timing cell which then readied the 
sensing photocells to examine many data 
columns simultaneously. 

(C//SI//REL) There were to be at least ten 
data columns, thus letters, packed into a linear 
inch of tape. To accommodate Wenger's need for 
counting more than single coincidences, ten let- 
ters were to be read at one time. This called for 
ten photocells for message characters, one to each 

(U) The engineer had to create a mask to 
ensure that light that shone through the first tape 
did not drift before it fell on the lower one. He 
also had to find a lens that would direct the light 
beams from overlapping holes, one for each col- 
umn, onto the correct sensing photocell. An allied 
problem was more challenging: he had to keep 
light from a coincident column from spilling over 
into the area of another column's photocell. The 
state of photocell technology did not allow easy 
solutions to any of the reader's problems. Among 
other problems, they remained fairly large. As a 
result, the young MIT engineer could not put ten 
of them directly under the columns of the 
Comparator's tapes. They had to be placed far 
under the reader and were arranged in a "U" pat- 
tern. That meant that the straight, parallel light 
from the coincident columns had to be accurately 
deflected. Moreover, complete electronic pack- 
ages for the photocells were not supplied by man- 
ufacturers. The MIT engineer had to tune each 
photocell and build the amplification circuits to 

turn the signals from the photocells into the dis- 
crete pulses needed by the third major compo- 
nent of the Comparator, the electronic counters. 

(U) The Most Difficult Problem of All, But It 

(U) With the knowledge of the tape and pho- 
tocell systems, the third young man began his 
work on the final details of what everyone 
thought would be the most difficult part of the 
project, its electronic counting system. 

(U) Precise digital counting with electronics 
was in its early years, and all attempts at creating 
tube-based calculating circuits were risky. 
Electronic tubes were designed for analog work, 
and it was only empirical tweaking that allowed 
them to be on-off switches. As late as 1940, the 
best experimental electronic counters worked at 
20,000 decimal counts a second during their 
cooperative periods. 

(U) One of his greatest challenges was the cir- 
cuitry for the Comparator's parallel-processing 
feature. It was needed to allow the machine to 
perform the simultaneous multiple letter tests 
that were so valuable to the cryptanalysts. 
Without parallel processing, the machine's power 
would be reduced by a factor of four. The student 
engineer had to construct five independent elec- 
tronic counters which were to tap the data from 
the reading station at the same time. The young 
man took the safe technological route, choosing 
to stay with the predictable and familiar gas-filled 

(U) The choice of architecture for the counters 
was also driven by the need to send the navy at 
least a feasible design, if not a machine, by mid- 
1938. Like the other electronic computer builders 
ofthe era, the young MIT engineer decided to 
imitate mechanical calculating machines. 41 His 
counters were decimal, not binary. Although such 
a design limited the range ofthe application of a 
computer, it was known to work and was simpler 


Page 41 


to construct than binary circuits. Each of the dec- 
imal counters was to consist of three or more 
rings often tubes with the needed electronics for 
arithmetic carrying, power, and control. 

(U) Providing the option of performing sever- 
al different analyses at one time meant additional 
challenges. Bush had designed the machine to 
allow the analysts to select the particular tests for 
each run. To permit this, the young engineer 
incorporated a set of "and" circuits that could be 
set to test for the desired combination. The 
Comparator's Rossi "and" circuit was the key to 
the machine's flexibility and parallelism. 

(U) In addition to the counters and the "and" 
circuits, the third engineer was handed another 
tough job. He was given the responsibility for cre- 
ating the banks of electrical relays needed to 
stand between the high-speed tube counters and 
the much, much slower printer. At the end of each 
pass, the counters had to be polled for their con- 
tents and numbers sent to the relays. The relays 
worked as a short-term memory, sending pulses 
to the magnets that controlled the print bars. 4 " 

(U) The Easiest Becomes the Most Difficult 

(U) There was a fourth man. He was in charge 
of the crucial data-entry system. The punch for 
the data tapes proved to be the Achilles heel of the 
Comparator. The problem was a perhaps 
inescapable result of the use of paper tape, as was 
Bush's inefficient 1 of 26 coding scheme. 

(U) The technology of the 1930s led him to 
reject a method of coding that could have 
increased densities on the tapes by at least a fac- 
tor of five and that would have led the 
Comparator's codes to fit with the navy's modern 
communication system. The use of a five-field 
character code, the Baudot code, would have 
allowed at least five letters to be placed on a line 
(column) of the 70mm tape. But the size and sen- 
sitivity of holes and photocells, the problems of 
aligning tapes, and the desire to limit the elec- 

tronics of the machine precluded the use of that 
coding pattern. 43 Bush's special coding scheme 
demanded a custom-made and very complex 

(U) An MIT machinist was instructed to make 
a keyboard-operated device to simultaneously 
punch two exact copies of a message. It had to 
keep the two tapes in perfect synchronization and 
to make precisely spaced tiny holes in each col- 
umn and row. The punch had to advance the 
tapes with absolute precision. Most challenging, 
it had to maintain the integrity of its tiny and 
sharp needle-like punching arms despite the 
impact as the arms struck the Eastman tape. The 
machinist was asked to devise tape cutters and 
the means to ensure that the spliced ends of the 
tapes would not pull apart during the runs. 
Unfortunately, the punch was the last component 
of the Comparator to be turned over to the proj- 
ect manager and then it was "not satisfactory." ** 
The punch's inadequacies cannot be blamed on 
the machinist; the responsibility has to be placed 
on the original design for the Comparator. 
Between 1938 and 1945 several teams of engi- 
neers tried to produce a viable data entry system 
for the paper tapes; none was able to build a 
rugged and reliable punch. 

(U) Beyond Murphy's Law 

(U) When Waldron MacDonald arrived in 
September, three student engineers had already 
sent their work to local machine shops. Bush 
trusted their judgment so much that, without 
examining the parts, he put MacDonald to writ- 
ing the descriptive reports for the navy. 
MacDonald took Bush's first schemes for each 
component, added what the students had done, 
and sent the reports to the navy for payment. 45 
The reports, including the final one submitted in 
the spring of 1938, were upbeat and gave the 
specifications for what everyone thought would 
be the first operating electronic data processing 
machine, 46 

Page 42 



(U) Although the reports contained a bright 
picture, the Comparator project had fallen victim 
to a host of problems. But the main reason for the 
problems in 1937 and 1938 was the technologies 
Bush so admired. They were not ready to be 
turned into useful machinery. Unfortunately, the 
results Bush and his young men expected on the 
basis of their early bench tests did not carry 
through to the parts they gave to MacDonald. The 
Comparator was far from ready for assembly. 
And only MacDonald was left to rescue it! 
MacDonald had much, much more to do than 
simply link the components together. Almost 
every component had to be reworked. 

(U) MacDonald put much thought and energy 
into reshaping the electronic components, and he 
more than fine-tuned the tape transport. More 
basic work had to be done on the reader. The opti- 
cal system needed a complete overhaul, and it 
took much of MacDonald' s attention. To bring 
the correct amount of light to each of the ten cells, 
he devised a 1930's version of fiber-optics. 

(U) Thus, MacDonald's assignment turned 
into something much more demanding than 
either he or Bush had imagined in mid-1937. 
MacDonald was not sure that he could solve all 
the problems of the transport, counters, and opti- 
cal sensors. Then chance compounded an already 
difficult situation. In a friendly game of touch 
football, MacDonald was knocked out by an 
unlucky "poke on the jaw." MacDonald remained 
unconscious and confined to bed for several 
weeks. His energy was seriously drained for 
months afterwards. 47 Despite the injury, Bush 
chose not to replace MacDonald. 

(U) What Hooper had complained about for 
so many years, the lack of appreciation of science 
in the navy, again struck the Comparator. 
Wenger, the strongest voice for a revolution in the 
technologies of signals intelligence and crypt- 
analysis, readied himself to leave for sea duty in 
mid- 1938. Wenger had to spend the five months 
before he was rotated putting the finishing touch- 

es on the detailed reports for Hooper's grand pro- 
posal for a modern communications system. 
Wenger left the country just a month before 
MacDonald shipped the troubled Comparator to 

(U) In spring 1938, MacDonald began test 
runs on the rebuilt parts. 48 He also had the chore 
of instructing the engineer the navy sent to learn 
about the machine. Wenger had arranged for a 
bureau technician to spend some time at MIT. 
During the spring, Frederick Dulong, one of the 
many ex-navy men who stayed on in Washington 
as civilian employees, was sent to MIT. 

(U) Wenger considered Bush very generous 
for having constructed a machine and approved 
Bush's suggestion that MacDonald be hired by 
the navy to fine tune the Comparator once it was 
in Washington. The bureau agreed and requested 
MacDonald to travel to Washington with the 
Comparator and to stay for three months. He was 
to adjust the machine and to instruct both techni- 
cians and cryptanalysts in its use. Safford, now in 
charge of the Comparator, was pleased that the 
bureau promised to give him some additional, if 
not permanent, help. 

(U) As soon as Bush signaled that a machine 
would be sent to Washington, Wenger began 
expensive preparations. He requested the money 
for tapes and lights and extra tubes, and he read- 
ied an area for the Comparator within OP-20-G's 
secret rooms. In a few weeks, additional funds 
were requested for the hardware necessary to 
prepare the tapes for the Comparator. 49 Wenger 
went much further. Describing a new era in crypt- 
analysis, he convinced the navy brass to give seri- 
ous consideration to funding more devices. 50 By 
the end of 1938, OP-20-G's budget request 
included more than $20,000 for additional Bush 
devices and special additions to the first 
machine. 51 In addition, "G's" new war plans con- 
tained a request for a Comparator for the pro- 


Page 43 


posed major cryptanalytic station at Pearl 
Harbor. 52 

(U) Spring Is a Time for Love, Not 

(U) When the Comparator arrived in 
Washington in late June, a month late, it would 
not start, 53 As bad, two of its most important 
parts had not been shipped -the punch and 
printer. About a month behind schedule and still 
only "semifinished," it found anew and well- 
intentioned guardian. But Fred Dulong could not 
give full attention to the machine. By mid-July, 
Dulong was able to run the counting circuits, 54 
but any more work was stalled because of the 
missing punch and printer. Unknown to anyone, 
they had been placed in a Cambridge safe-deposit 
box byMacDonald to await his return to the 
country in August 55 following a honeymoon. 

(U) The cryptanalysts certainly did not have 
the time to wet-nurse the Comparator. While the 
bureau's men bewailed the results of becoming 
entangled with an impractical professor, the 
cryptanalysts in charge of the day-to-day work 
were coming under incredible pressures to pene- 
trate all of the sophisticated Japanese code and 
cipher systems. Japan's invasion of China in 1937 
had made it clear that war was imminent, 56 and 
by 1938 OP-20-G was facing crisis conditions. 
The sinking of the Panay in December led to a 
scramble to protect American codes. In addition, 
there were hints that Japan was about to make 
another sweeping change in its codes and to 
introduce its Purple cipher machine. 57 What 
energies OP-20-G had were necessarily devoted 
to developing techniques and machines that gave 
immediate results. Its faith was, quite naturally, 
placed in the direct analogs of Japan's encipher- 
ing machines, and its men wanted resources 
devoted to modernizing the tabulators. 

(TJ) Thus Waldron MacDonald did not arrive 
in Washington at the right time for any experi- 
mentation at"G" or the bureau. Driving from 

Cambridge in August 1938, he had the 
Comparator's punch and printer in the back of his 
station wagon. Working in OP-20-G's downtown 
offices, MacDonald attempted to save his and 
Bush's reputation. 

(ID He hurried the Navy Yard's effort to build 
tape duplicators and splicers and soon convinced 
the bureau to build a new punch. The one from 
MIT could not be coaxed into working. Don Seller 
took on that challenge. 58 Then MacDonald began 
working on the other components. Although no 
major changes were made to the Comparator, it 
took an unexpected fourth month of work to 
announce a finished machine in November. 

(U) In late 1938, OP-20-G's leader, Safford, 
congratulated Bush and informed him the crypt- 
analysts and the bureau's men planned to spend 
the next year experimenting with the wonderful 
and reliable machine. Possibly because they now 
realized how much a well-schooled optical elec- 
tronics engineer would cost, OP-20-G did not 
make an effort to hire a replacement for the MIT 
engineer or, as planned earlier in the year, to con- 
struct at least one more Comparator. 59 

(U) R.AM Project Seems to Die, Late 1938 

(U) With Wenger gone, no one pressed for an 
immediate extension of the program. 60 Bush, 
in turn, quickly fended off another attempt by the 
navy to link him to "G's" projects. The conse- 
quences of the failure to continue on with 
the Comparator project in 1938 were severe. 
Soon after MacDonald left Washington, the 
Comparator again became inoperable. It was so 
temperamental that the only attention it received 
was from Dulong, whose many other duties 
allowed just part-time work. 61 It was listed on 
OP-20-G's equipment roster in 1939, but it was 
never used, not even on the type of 
important project for which it had been designed, 
the breaking of the Japanese Purple cipher 
machine. 62 Its technical problems become so 
great that it was removed from the cryptanalysts' 

Page 44 



quarters and sent to the Navy Yard where it could 
be tinkered with. 

(U) Although overworked because of the 
Japanese code and cipher crises, Safford had 
asked for a report on the Comparator and 
received some very disheartening news. Dulong 
responded that nothing but the electronic coun- 
ters proved reliable, and the machine had not 
been functional long enough to allow in-depth 
development of procedures. The Navy Yard's men 
did not think there was any possible quick fix for 
the device. Most ominous was the failure of the 
data entry component, the punch. Even the sec- 
ond version of that purely mechanical and sup- 
posedly simple mechanism could not be made to 
produce precise tapes. There was little hope of 
basing an entire system of analytical machines 
around the original Bush design if there was not 
an efficient and reliable data entry device. 63 In 
1940, Safford, who two years before declared the 
Comparator a reliable and useful invention, had 
to admit the machine never worked and that the 
entire project had not progressed as planned. 

(U) A Comparator There May Never Be 

(U) In late 1940 Bush gained another chance 
to prove the power of optical-electronic machines 
and the ability of academics to create the tech- 
nologies of defense. 6 * He arranged for MITs 
John Howard and his men to rescue the first 
paper tape Comparator and to design the long- 
promised microfilm version. 

(U) This second MIT OP-20-G project of late 
1940 is of extreme historical importance because 
it became the foundation for the United States 
Navy's incredible Rapid Machines Program of 
World War II. That little known adventure rivaled 
Britain's famous work on the Bombes and the 

(U) Tragically, that program is also important 
because of its failures. Although it began with 
expectations of producing electronic digital 

machines to attack the feared cipher devices of 
the Axis powers, it turned to older technology and 
logic. To be able to provide anything of value to 
OP-20-G, Howard's men had to step back from 
electronics, digital techniques, and microfilm. 
Although the navy's cryptanalysts began World 
War II with promises that electronics could be 
made to work, they had to wait for almost two 
years after Pearl Harbor before any machines 
appeared that affirmed that Bush's ideas had 

(U) The story of John Howard's navy project 
has to begin with the crises in Europe and Asia, 
policy decisions in the White House and London, 
and the organization of American science in 
World War II. 

(U) Big Science Begins to Emerge 

(U) Bush's high-science friends were active in 
more than the cause of research. They were 
among the nation's earliest supporters of a posi- 
tive response to the German threat. They lobbied 
for the creation of the National Defense Research 
Committee (NDRC). The NDRC was the realiza- 
tion of Bush's ideal of how to link academia and 
the military. Given almost complete power by 
Roosevelt to shape the NDRC, Bush laid down 
ground rules that gave power to academics to 
begin research projects and to be free of military 
control. Having its own funds and being a presi- 
dential creature, the NDRC and its more powerful 
extension, the Office of Scientific Research and 
Development, could initiate blue-sky programs 
and carry them through to development. 

(U) One ofthose programs interlaced the 
NDRC with American cryptanalysis, but only 
after it had dealt with along list of projects of 
much higher priority. Atomic power and radar 
were the leading problems, and the scientists at 
the most prestigious universities and corporate 
research centers received the first calls from the 
NDRC's leaders. 


Page 45 

TOP OCOrcrn/OO M I N TOnCL TO USA, AUg , CAN, GRP nnri N7U/Y1 

(U) The executives at the NDRC realized that 
atomic research and the development ofthe 
potentials of radar called for advanced computa- 
tion, but, alone, those problems would have led to 
a minimal NDRC involvement in computers. It 
was a lower priority challenge that plunged the 
NDRC into computer research and established 
who would participate in the navy's future Rapid 
Machine effort. Atomic scientists were calling for 
electronic control devices, but most important for 
the history of OP-20-G was the hope that radar 
could be used to automatically control antiair- 
craft weapons. That led to the NDRC's involve- 
ment in the development of electronic fire control 
computers in the early 1940s. 65 

(U) The exploration of such electronic digital 
machines was the perfect type of work for the 
NDRC because it centered onunproven and 
experimental technologies. The NDRC's scien- 
tists believed that digital electronics had poten- 
tial, and they rekindled the fire control projects. 
Hundreds of thousands of dollars were poured 
into fire-control computer and atomic-counter 
work in the first two years of NDRC's life. 

(U) Fire Control 

(U) The NDRC began the first stages of its fire 
control project in June 1940. Bush's old friend 
Warren Weaver ofthe Rockefeller Foundation 
assumed command. The research at RCA, which 
had led to the design ofthe fastest binary circuits 
in the nation, if not the world, was picked up by 
the NDRC. Then Weaver coordinated the work at 
RCA with wide-ranging explorations at Eastman, 
MIT, Bell, and, to some extent, NCR Of signifi- 
cance for the history of OP-20-G's machines, IBM 
was again left out ofthe NDRC circle although its 
centers of electronic research were working on 
quite advanced components and systems. 66 

(U) Because ofthe NDRC's stimulus, by the 
time of America's formal entry into the war, RCA 
Eastman, Bell, and MIT had several proposals for 
digital-based fire control systems, ones the NDRC 

evaluators thought had great promise. In the 
spring of 1942, meetings were called, and all par- 
ticipants shared their knowledge and designs. 67 
The reports ofthe fire control projects were made 
available to the American technical community, 
which now included John Howard. He was made 
aware ofthe designs for the most advanced com- 
puter components. 

(U) Many ofthe fire control developments 
would find their way into cryptanalytic machines 
and into such pathbreaking computers as the 
ENIAC. By mid-1942, there were great hopes for 
the development of at least a prototype electronic 
gun controller. But Warren Weaver and his assis- 
tants concluded that digital electronics was too 
good. It was too fast and too precise for the guns 
used by the military. In July 1942 the fire control 
program was dropped - but with three important 
exceptions. The development projects for the 
Eastman film-based analog-to-digital signal con- 
verter and RCA's fabulous multifunction 
Computron tube were to be continued, as was 
NCR's counting circuit research. Although they 
were viewed as long-term projects, the three 
efforts were financed for only a few more months 
because the press of other work forced the NDRC 
to abandon them. 68 

(U) The Second Comparator 

(U) Meanwhile, just weeks after the work on 
high-speed electronic counters and fire control 
computers had begun, Bush and OP-20-G came 
together. A visit with Bush in early summer 1940 
indicated a reawakening of interest in the original 
Comparator, which had sat unused at the Navy 
Yard for almost two years. But it was not until 
October 1940 that anything was done about its 

(U) A limited and secondary role for MIT was 
unacceptable to Bush, however. He returned to 
his old demand for freedom from bureaucratic 
control, and, within a few weeks, he was able to 
reshape the first murmuring about a new 

Page 46 



Comparator into a project that satisfied his ambi- 
tions. Bush wanted a prototype of a microfilm 
Comparator. While the first Comparator would 
continue to be a paper-tape, the second genera- 
tion Comparator was to be centered about micro- 
film. Bush soothed Laurance Safford's anxieties 
about optical and electronic machines and told 
him that the new microfilm version of the 
Comparator would be delivered in time to be of 
use in the coming war. In late 1940, Safford 
encountered litde resistance to the idea of trans- 
ferring the project to MIT. The navy's cryptana- 
lysts were too busy battling the Japanese naval 
code, and too worried about taking on the 
German systems to care about the loss of control 
over unusable machinery. 

(U) 0P-20-G and Ultra 

(U) The navy's bureaucrats were less happy 
about yielding control. They agreed to most of 
Bush's conditions although they did want a con- 
tract and agreed to have the MIT work coordinat- 
ed through the navy's Office of Research and 
Inventions. The BuShips' (Bureau of Ships) 
demands in late 1940 were much less severe than 
in 1935, but it took some additional political 
maneuvering to put the Rapid Machine project 
back into the hands of the MIT students. 

(U) Under the 1940 agreement with the 
Bureau of Ships, Bush had full control of the new 
Comparator project, and his men were to work at 
MIT, not at the Navy Yard. The navy also agreed 
to wait for the results of the new MIT work before 
considering the construction of any more Rapid 
machines with or without MIT involvement. 

(U) As early as mid- 1940, the most important 
Americans were informed of some of Britain's 
promising though still limited powers over a few 
German cryptologic systems. But OP-20-G was 
not told how to break the Enigma or other impor- 
tant ciphers. Despite the British promise to share 
the information from Ultra, the Americans feared 
a British monopoly over Enigma. In addition, in 
early summer 1940 there were fears that Britain 
would collapse. OP-20-G's cryptanalysts worried 
they would have to assume responsibility for 
Enigma, something for which they were totally 
unprepared. 69 

(U) The U-boat threat had already led to 
British pleas that OP-20-G and Naval Intelligence 
shift their scarce resources to direction-finding 
and traffic analysis to compensate for their inabil- 
ity to read any significant German naval system. 70 
The cryptanalysts in Washington thus had little 
time to waste on what some of them regarded as 
Bush's technological fantasies. The navy's engi- 
neers, already overworked creating analogs of 
encryption machines, building advanced radio 
equipment, and helping to revise OP-20-G's tab- 
ulators, were happy to be rid of the "college pro- 
fessor's" folly. 

(3//3I//REL) OP-20-G may not have told 
Bush, but it was not relying solely upon his ideas 
for machine processing; it had to protect itself 
through the use of older and proven technologies. 
The experienced cryptanalysts had insisted on a 
tabulator program, one that was to remain under 
their direct control. MFTs men were to have no 
say about the new tab projects. In early 1941 IBM 
was contacted about making the major changes in 
its machines required to allow its equipment to 
perform new tests. For example, relays were 
added to the machines to strip superencipher- 
ments from the Japanese code and to flag repeti- 
tions of code groups. By mid-year IBM was asked 
to do much more and to give OP-20-G very spe- 
cial attention. 71 

(U) Bush was finally able to circumvent the 
bureaucracy and go his own way, perhaps with 
ambitions to create a full Rapid Machine center at 
the Institute, one free of military interference. 72 
The old Comparator was to be shipped to MIT for 
repair, and a new one was to be designed and con- 
structed in Cambridge. 

(U) Howard quickly became the man in 
charge of the Comparator project. Although the 


Page 47 


(S) 70mm Comparator 

future of Bush's ideas rested upon the new micro- 
film Comparator, the old paper tape machine and 
its punch became the focus of attention. The 
punch was a critical problem because its two pre- 
vious versions were failures. Understandably, 
Howard urged his men to use caution as well as 
creativity, but the slow tempo of his project soon 
generated concern within the Bureau of Ships. 
During the first months of 1941, as time passed 
without results, the navy found it more and more 
difficult to accommodate having its project run by 
a civilian agency. 

(U) So Long for So Little 

(U) It took almost a full year to redo the old 
Comparator, the only Rapid Machine. It arrived 
in Washington three weeks after the Japanese 
struck at Pearl Harbor. 73 It may have taken longer 
than expected to deliver the old Comparator 
because of adding one new feature to Bush's 1938 
design. The "locator" performed a function the 
navy had thought of adding in the late 1930s. It 
allowed the use of a transverse tape to find more, 
complex pattern matches than was possible with 

the original system. With the locator, the code- 
breakers could quickly identify which messages 
held important code or cipher groups. 74 

( S ) In the private language of the codebreak- 
ers, "locating'' was a Brute Force approach to 
finding possible "depths." 

(U) Once inaction in mid-1942, the old 
Comparator did help crack the Japanese naval 
attache cipher machine, but the Comparator's 
newest punch also malfunctioned. 75 In addition, 
the machine's bad temper called for a visit by one 
of the MIT men, Larry Steinhardt, who had to 
simplify the device to achieve reliability. 76 

(U) The Search for the Second Comparator 

(U) While struggling with the old 
Comparator, the young men at MIT paid atten- 
tion to its new microfilm version. In early 
November 1941, the MIT-NDRC group was so 
positive about the future of a microfilm 
Comparator that Bush obtained another signifi- 

Page 48 



cant grant from the NDRC. 77 Then, although the 
NDRC was not supposed to be involved in pro- 
duction, Howard awarded a $25,000 contract to 
National Cash Register's electronics laboratory. 
Joe Desch agreed to build as many as thirty 
copies of the sets of new high-speed counters and 
fast printers needed for the future microfilm 
Comparators. 78 

(ID In a November 1941 meeting, it was 
declared that construction was ready to begin on 
the next Comparator. Expecting to see the newest 
Comparator in a few months and viewing the MIT 
group as a long-term resource, the OP-20-G ana- 
lysts outlined needs for other devices. One of 
those outlines had a hidden significance. It would 
connect the MIT men to the Ultra Secret although 
they did not know of Britain's ability to crack the 
Enigma nor of the critical negotiations between 
OP-20-G and Britain over sharing intelligence 

(U) Not Equal Partners in Ultra 

(U) Agreements were made at the very high- 
est levels in 1940 and 1941 for Britain and 
America to share cryptanalytic methods as well as 
the military information that came from signals 
intelligence. However, England's wizards did not 
have a mathematical solution to the Enigma! 
Without good guesses as to key words in mes- 
sages and knowledge of the inner workings of the 
Enigma radio networks, Britain could, and 
would, become deaf. 79 

naval Enigma. However, they made little progress 
toward what the Americans needed, a purely 
mathematical cryptanalytic solution. 

(U) Although she had helped break into simi- 
lar devices, was informed of some of the British 
methods, and labored for almost a year, Driscoll 
could not find the ways and means for an 
American Enigma solution. Fortunately, she was 
willing to ask for help. During the November 1941 
meeting between MIT and OP-20-G, she 
described her needs, and Howard was asked to 
think of ways to automate 
her "problem." She was 
determined to devel- 
op a method more 
permanent than 
the ones Britain 
had chosen. 
that called for 
a machine 
somewhat dif- 
ferent from the 
accepted the 
and Driscoll was 
happy with the 
promises by the 
young men from 
MIT. 80 

O Agnes Meyer 

(C//KF . I )) Perhaps it was the fragility of the 
solutions that made the British somewhat less 
willing to share their secrets with the Americans. 
Whatever the reason, the Americans began to 
think it was necessary to have their own anti- 
Enigma capability. In late 1940, OP-20-G shifted 
their one professional cryptanalyst, who had just 
made the first entries into the Japanese fleet 
code, to the German problems. The venerable 
Agnes Meyer Driscoll and three young navy offi- 
cers began an attack on the frustrating German 

(U) Another Machine That Wouldn't 

(U) The cordial meeting with Howard in early 
November 1941 impressed OP-20-G's people. But 
OP-20-G and the Bureau of Ships became very 
worried and skeptical about university work 
when, just a few days after the Washington con- 
ference, 81 Howard wrote the navy that experi- 
ments were showing the new Comparator's 
microfilm to be deforming when used in test 


Page 49 


assemblies. The navy must have wondered how it 
could have taken the MIT group so many years to 
discover its primary technological assumption 
was untenable. They may also have asked how 
OP-20-G's need for revolutionary cryptanalytic 
devices could be fulfilled if the responsibility con- 
tinued to be left in the hands of the inexperienced 
NDRC and the young MIT students - people who 
failed to test underlying assumptions before wast- 
ing a critical year's work. 

(U) The Revenge of Mechanics: 

the First 

(U) While John Howard had been facing up to 
the failures of photo-optics and electronics, some 
practical men were creating immediate "mechan- 
ical" solutions to cryptanalytic problems. 
Beginning in late 1940, the engineers in the 
army's and navy's cryptanalytic branches began 
to work closely with IBM and its engineers. The 
outcome was the first operational special-pur- 
pose cryptanalytic machine and the first of a long 
and important series of modifications to IBM's 
standard offerings. 

(TD//0I//REL) By spring 1941, the army 
had its Gee Whizzer working on the transpo- 
sition ciphers of several nations, and the navy 
was about to receive the first of its special 
IBM Navy Change Machines. 82 

(TS//SI//REL) Logs and Relays 
Gee Whizzer 


(TS//SI//PJEL) The Gee Whizzer had 
been the first to arrive. In its initial version it 
did not look impressive; it was just a box con- 
taining relays and telephone system type 
rotary switches. But when it was wired to one 
of the tabulating machines, it caused amaze- 
ment and pride. Although primitive and ugly, 
it worked and saved hundreds of hours of 
dreadful labor needed to penetrate an impor- 
tant diplomatic target. It proved so useful 
that a series of larger and more sophisticated 

"Whizzers" was constructed during the war. The 
last of the four versions had an electronic matrix 
and was in operation throughout the decade. The 
navy admired the Whizzer so much that it built its 
own version, the Jeep. 83 

(T0//3I//REL) The Gee Whizzer was born 
because of a specific problem that arose in early 
1941. It was the Japanese diplomatic service that 
had caused the SIS group to search for a new type 
oftechnological solution. When the Japanese 
made one of their diplomatic "transposition" sys- 
tems much more difficult to solve through hand 
anagramming (reshuffling columns of code until 
they made "sense"), the American army did not 
have the manpower needed to apply the tradi- 
tional hand tests. 

(Tfl//3I//REL) Friedman's response was to 
try to find a way to further automate what had 
become a standard approach to mechanically 
testing for meaningful decipherments. His search 
did not include electronics. Rather, he told Leo 
Rosen to find quick ways to extend the power of 

(TS//SI) Gee Whizzer 

Page SO 



the IBM machines that were beginning to arrive 
at his offices in greater numbers. Rosen's first 
task was to learn how automatic anagramming 

(TS//SI//RFI1) One of the most traditional 
ways of hiding the plain language or even the 
codes in a message was to transpose the columns 
of the text. With columns moved around in a ran- 
dom way, it was very difficult for those who might 
intercept a message to realign the text to its orig- 
inal order. The old hand attack had been to move 
one column after another against each other with 
an analyst making continuous judgments as to 
whether the new alignments were building 
towards a meaningful plain language arrange- 
ment. That was a tedious and time-consuming 

(T9//SI//PFI ) During the 1930s the SIS had 
made some progress towards easing the analysts' 
burdens. Statistical studies of various languages 
had been made and a system of weights had been 
calculated. Turned into "logs"(logarithms) so that 
addition rather than multiplication could be used 
to build scores for combinations of letters, log 
weights were assigned to each letter in a trans- 
posed message. As each of the columns was 
rearranged, the weights were summed and an 
evaluation was made as to whether the sum 
approached that expected for a column of plain- 
text. If the logarithms of the statistically expected 
frequencies of the combinations were high, it 
indicated that the correct order of the text 
columns had been found. The results were dou- 
ble-checked by an analyst to see if the realigned 
columns made plain-language sense. 

(TS//ST//FFL) The log weight method had 
been implemented on the tabulating machines, 
but the process entailed much special card- 
punching and many runs of the cards to align all 
the columns. Worse, the tabulator method did 
not include an automatic test for plain-language 
"build-up." That meant that bad column 
sequences might be run for too long and, worse, 

all results were printed out. All of those usually 
worthless printouts had to be examined by an 

(U) The method worked, but it was very, very 
labor intensive even with the use of tabulators. It 
took too much time to feed the round after round 
of cards that were required to test all columns of 
a transposed message against all others. 

( T0//0I//REL) Rosen and the IBM consult- 
ants realized that not much could be done about 
the cards; there was no other viable memory 
medium. But it was thought that it might be pos- 
sible to eliminate all but significant results from 
being printed. Rosen and his men, with the per- 
mission and help of IBM, turned the idea into the 
first and very simple Gee Whizzer. The Whizzer's 
two six-point, twenty-five-position rotary switch- 
es signalled the tabulator when the old log values 
that were not approaching a criterion value 
should be dropped from its counters. Then they 
instructed the tabulator to start building up a new 
plain-language indicator value. 

( T S //SI//REL ) Simple, inexpensive, and 
quickly implemented, the Gee Whizzer reinforced 
the belief among the cryptoengineers in 
Washington that practical and evolutionary 
changes were the ones that should be given sup- 

(U) The Navy Gets Some Changes 

(U//POUO) OP-20-G's enlisted grade in- 
tra use engineers felt the same way and argued for 
the help they needed to turn their imaginative 
ideas for true cryptanalytic machines into hard- 
ware. Their requests reached the office of the 
Director of Naval Communications and in mid- 
1941 Captain Redmond informed them that he 
had used his personal influence to get IBM's Tom 
Watson to agree to help the navy. 

(U//FOUO ) An IBM executive immediately 
came to Washington, listened to the ideas of Pete 


Page 51 


Deffert and Lou Holland, and gave his blessing to 
their hopes for advances more radical than just 
attaching relay boxes to standard machines. An 
IBM engineer was soon assigned to duty at "G," 
and he began to refine the navy engineers' sug- 
gestions and to forward them to IBM's designers. 

(TS//SI//REL) Within a few months the first 
Naval Change (NC) was up and running. The NC- 
1 automatically sensed the beginning of a series of 
cards and then punched an increasing serial 
number in each successive card. IBM delivered a 
more complex machine just a few weeks later. 
The new automatic cross-footer also worked from 
the day it was installed. It provided a high-speed 
means of decrypting additive cipher systems, 
such as those used by the Japanese Navy. 

(U) The NC series was continued throughout 
the war. The thirteen different machines became 
progressively more complex, but each worked, 
and none were burdens to the maintenance engi- 
neers at M G." 84 

(U) The Greatest Kludge of. All, But It 

(0//GT//REL) The navy's enlisted men were 
involved in something more ambitious: the con- 
struction of the mechanical contraption that 
worked, the M4. The Washington Code and 
Signal Section's electricians and machinists put 
fifty wheels, each having thirty contacts and ten 
stepping notches, together with ten banks of 
lights and a set of hand cranks, to provide an 
automatic way to identify what additives had 
been used in Japanese messages. The machine 
exploited a weakness in Japanese systems -all 
the code groups had to be divisible by three. 

(S//SI//RBL) To find a likely additive, ten 
code groups were set on the machine; then the 
additive was entered with the cranks, and finally 
the machine was ordered to find out how many of 
the resulting deciphered code groups were divisi- 

ble by three. The lights told the operator which 
groups were "over," "under," or "divisible." m 

(U) Trying to Save Bush's Reputation 

(U) John Howard was probably not told of the 
triumphs of the practical navy engineers and the 
old technologies, but he knew that he had to do 
something to save Bush's dream. He came up 
with a very rudimentary substitute for the film 

(U) He advised the bureau that photographic 
plates could be substituted for microfilm. 86 
Although very pessimistic, Howard did not give 
upon the Comparator entirely. But the new 
Comparator project seemed to be another very 
embarrassing disaster. The bureau certainly was 
unhappy, and the navy's cryptanalysts thought 
they might be left out of the electronics revolu- 

(U) Bush was upset that his plans for elec- 
tronic cryptanalysis were in trouble. There was 
almost nothing to show for a decade's work. And 
John Howard's bad news could not have come at 
a worse time for the navy. He made his confession 
about the microfilm Comparator just as the 
American intelligence agencies were frantically 
searching for the final clues to where Japan 
would attack. In a few weeks OP-20-G had to face 
the consequences of the failure to predict Pearl 
Harbor. But a combination of factors gave Bush's 
young men yet another chance. The ability of 
Howard to continue on independently (because 
he had a year's NDRC funding remaining) was 
important, but a more significant reason was the 
combination of the return of Joseph Wenger and 
the political influence of Vannevar Bush. 

(U) Yet Another Chance 

(U) Wenger returned from sea duty in the 
summer of 1941. Although assigned toOP-20's 
war plans section, 87 he contacted the cryptana- 
lysts and Bush about the outcome of the year of 

Page 52 



NDRC work. After hearing of the situation, and 
despite Howard's bad news, Wenger talked with 
his contacts at OP-20-G and pleaded for a contin- 
uation of the relationship with MIT. 88 His urging 
and the navy's dread of alienating the head of the 
NDRC, Vannevar Bush, allowed Safford to begin 
a program that would be vastly expanded when 
Wenger was sent back to OP-20-G in early 1942. 

(U) When the Ciphers Can't Be Broken 

(U) Wenger's influence at OP-20-G was the 
result of his long involvement in modernizing 
naval communications. He had a reputation as an 
expert in all communications fields. He was 
America's leading advocate of a high-tech alter- 
native to cryptanalysis. In the 1930s, Wenger pre- 
dicted that unless massive breakthroughs were 
made in cryptanalysis, such as the construction of 
a full range of Rapid Machines, it would be fool- 
ish to rely upon direct communications intelli- 
gence such as codebreaking. Until America built a 
truly innovative mathematical cryptanalytic 
capability, he argued, other signals intelligence 
resources had to be exploited. Wenger argued 
that codes and ciphers were becoming too com- 
plex to crack with available techniques, and, as 
important, an enemy's frequent changes of sys- 
tems would always create blackouts at the most 
critical moments. 89 

(U) Wenger had become America's advocate 
for what became known as "traffic analysis." He 
had spent years studying and developing T/A In 
traffic analysis, the concern was not with the con- 
tent of messages but with the easily identified 
callsigns of senders and receivers, the timing and 
numbers of messages in a network, and the shifts 
in patterns of transmissions. 9 ° Although not as 
glamorous or exciting as cryptanalysis, traffic 
analysis was not a low-tech activity. 

(U) Many aspects of T/A called for more eso- 
teric and expensive hardware than traditional 
codebreaking. The method depended upon 
sophisticated direction finders to locate enemy 

stations and on other expensive radio equip- 
ment. 91 It was also very demanding in terms of 
personnel and data processing equipment. 

(U) The first step for T/A was the ability to 
intercept enough messages. Revolutionary auto- 
matic scanners searched for active channels, 
oscilloscopes helped identify stations and opera- 
tors, and very sensitive receivers plotted trans- 
missions. 92 The hardware was not the end of it, 
however. Optimal radio interception and plotting 
called for advice from physicists; the exploitation 
of the intercepts needed advanced statistical -ana- 
lytical techniques. The intercepts and location 
estimates had to be correlated and subjected to 
time-consuming analysis. The tabulators were 
frequently called upon to compile the necessary 
interaction matrices. The expense and manpower 
T/A needed seemed worthwhile. Wenger's recon- 
struction of Japanese naval maneuvers from T/A 
analysis during the mid- 1930s was a triumph. 

(U) By 1940 OP-20-G's intercept crews were 
logging thousands of messages a month from the 
Pacific and the Atlantic, and the method was con- 
sidered essential. The SIS had also begun to 
appreciate T/A and sent its top men to the Canal 
Zone and Hawaii to establish intercept and pro- 
cessing sites. Those investments were 
inescapable. With America and Britain unable to 
read the most important German systems, T/A 
was the only hope in the West. 93 

(U) T/A had its limits, however. It could not 
reveal long-term plans; it gave just a picture of 
immediate intentions. It had other imperfections 
as well. The most important was a dependency on 
very frequent communications. If a station did 
not broadcast, it could not be identified and locat- 
ed. Tragically, in 1942 T/A was unable to deal 
with the German submarine onslaught because 
the submarines off the American coast followed a 
routine of radio silence. 


Page 53 


(V) Wenger to the Rescue 

(U) Joseph Wenger's influence atOP-20-G 
was not diminished by the failure of T/A to live up 
to its promise. In 1942 he was granted the power 
he needed to implement his plan for a centralized 
organization for naval communications intelli- 
gence. Wenger's ideas were quickly accepted. 94 
Along with the approval of his plan came his 
appointment as the operating head of"G." In 
February 1942 he began to reorganize "G" and to 
revive Hooper's dream of bringing science and 
cryptanalysis together. 

(U) But those were difficult times for Wenger. 
In early 1942 the U-boats began to slaughter 
dozens of freighters insight of the American 
coast. The U-boat threat to the Atlantic convoys 
was growing so fast that a continuation of the 
sinkings threatened Britain's survival. Britain was 
unable to penetrate the new naval four-wheel 
Enigma, and OP-20-G remained without any 
power over German systems. 

(U) Mathematics to Meet the Great 

(U) The Atlantic crisis had a strange impact 
on OP-20-G's future. It both helped and hindered 
Wenger's crusade for the Rapid Machines. In its 
first phases, the crisis aided him. As part of his 
outline for the expansion of OP-20-G, Wenger 
had planned for the creation of a special research 
group. Its mission was to apply formal mathe- 
matics to cryptanalysis. OP-20-G had never 
before had a professionally trained mathematical 
team. In February 1942 Wenger brought together 
the few professional mathematicians who had 
already been called to service and had them join 
the T section. Of significance for World War II 
and the history of the Rapid Machines, they were 
handed more than mathematical responsibilities. 
They were ordered to take on some technical 
radio problems. In addition, they were given 
some of the responsibility for the critical German 

(U) Next, the "M" section was handed 
Wenger's pet, the Rapid Machines project. 
Luckily, Wenger found the right man, Howard 
Engstrom, to direct the third major attempt to 
make optics and electronics into cryptanalytic 
tools. When first called for the war crisis, 
Engstrom was asked to give advanced technical 
advice to OP-20-G's radio intercept and direction 
finding group. 95 That T/A assignment was a very 
important and demanding post. But heading "M" 
turned out to be much more of a challenge. The 
new job called for political as well as technical 
skills. By spring 1942 the Rapid Machines had 
again become political creatures. 

(U) The apparent failure of Howard's 1941 
microfilm Comparator had not ended his work or 
the interest of OP-20 in cryptanalytic machinery. 
But it did reopen the old battles over control of 
innovation in the navy. 

(U) Bureaucracy vs. Science, Again 

(U) Pearl Harbor, despite the blame hurled at 
the army and navy intelligence agencies, led to 
the release of funds and energies for cryptanaly- 
sis. For the first time in its history, OP-20-G had 
enough money to pursue technological dreams. 
In response, in early 1942, Safford, still the head 
of OP-20-G, initiated a survey of needs and wrote 
out a wish list that included Rapid Machines. 
There was enough money to explore all options. 
The first choice of the operating cryptanalysts was 
IBM electromechanical machinery. They asked 
for more standard equipment and for the devel- 
opment of a host of special attachments. Safford, 
not yet having an "M" section, and very short of 
personnel, turned to the Bureau of Ships for tech- 
nical and administrative help. 

(U) He found it easy to convince the bureau to 
deal with the trustworthy IBM. Very soon, the 
Bureau of Ships established what it saw as a har- 
monious three-way relationship between IBM, 
the old hands at"G," and itself. 

Page 54 



(U) In the first weeks of 1942 the bureau 
decided to allow OP-20-G to invest navy funds in 
an exploration of Rapid Machines, but it 
demanded a heavy price, one which included a 
radical change in the relationship between MITs 
men and the navy. The bureau's men, not those 
from OP-20-G, were to run the technical and 
financial parts of the program. Above all, the 
bureau wanted the projects out of the halls of 
MIT. Its officers demanded that all work be done 
by established corporations that followed the 
navy's standard procedures. But in early 1942 it 
began to take charge of the NDRC project, giving, 
it thought, badly needed managerial direction. 

(U) Meanwhile, the bureau explored ways to 
decrease its dependency on Bush's group. Then 
the bureau decided to show its power. It took the 
Comparator away from Bush and MIT. In March 
1942 Bush's structure for linking the military and 
academia, at least for cryptanalytic machines, 
began to be dismantled. The work at MIT, per- 
haps with the exception of the designing of a 
punch, was ended, and the secret workshop was 
shut down! 

(U) A Seeming Victory for Science 

(U) As soon as Wenger had returned to "G" 
and learned of the bureau's actions, he feared that 
the corporate projects would produce machines 
the cryptanalysts could not use. Wenger began an 
attempt to shift power back to OP-20-G. A search 
was begun for experienced engineers to augment 
the "M" group. By the end of 1942 OP-20-G had 
some of the leading men in computer electronics. 
Through formal and informal means, the name 
"M" came to mean machinery as well as mathe- 
matics. 96 

(U) Wenger convinced the bureau to give OP- 
20-G's Rapid Machine program near autonomy 
as well as its own facility and workforce. But it 
took a major intelligence crisis to achieve that. 
Wenger would not have been so successful and 
there would not have been a Naval Computing 

Machine Laboratory at the NCR factory in 
Dayton, Ohio, if the British had been able to con- 
quer the German submarine Enigma system or if 
the White House had insisted that OP-20-G 
remain dependent on Britain's Ultra. 

(U) The establishment of the Naval 
Computing Machine Laboratory and the 
increased power of the Rapid Machine group did 
help OP-20-G to build a series of innovative 
cryptanalytic machines, including the American 
version of the Bombe. By the end of the war, the 
American navy had some of the world's most 
advanced electronic machines. 

(U) Notes 

1. (U) A recent study of the messages intercepted 
by the Americans in 1941 concludes that if there had 
been the manpower to decode all the messages it 
would have been clear that Pearl Harbor was a target. 
Frederick D. Parker, "The Unsolved Messages of Pearl 
Harbor," Cryptologia 13(1991): 295. 

2. (U) Library of Congress, Papers of Stanford 
Caldwell Hooper: March 3, 1933 to January 1,1935, 
Correspondence with Redman and Jewett: "Contact 
scientists"; June 1, 1934, "Binaural Sons of C"; October 
20, 1935, "Travel to Laboratories"; Box 18, Hooper to 
Secret Naval Board, "Communications Plan," 
February 7, 1936; June 10, 1935, "McDowell, Contact 
scientists"; and November 20, 1935, "Travel to 
Boston." Navy Biographies Section OI-140, 27 April 
1945, "Rear Admiral Stanford C. Hooper, U.S. Navy, 
Deceased." NARA RG457, SRH-355, "Naval Security 
Group History to World War II," 269. 

3-(U) Note that Hooper's plan came almost a 
decade before the British began their now famous 
project atBletchley Park to develop automata for 
cryptanalysis. For an example of the results of his 
efforts to modernize OP-20-G, NARA RG457, SRMN- 
083, "Military Study of Secret Radio Calls," January 8, 
1938, by Joseph N. Wenger. 

4. (U) The navy did not include signals intelligence 
in its formal war plans until 1937. NARA RG457, 
SRMN-084, "The Evolution of the Navy's Cryptologic 
Organization," 3. On the attacks against military and 


Page 55 


diplomatic codes in the early 1930s, NARA RG457, 
SRH-159, "Preliminary Historical Report of the 
Solution ofthe 'B' Machine," 12. RG457, SRH-355, 
"Naval Security Group History to World War II," 82, 
NARA RG457, SRH-305, The Undeclared War; The 
History of RI," 15 November 1943, byLaurance F. 
Safford, Captain, U. S. Navy. 

5. (U) See Harold G.Bowen, Shy)s, Machinery 
and Mossbacks; The Autobiography of a Naval 
Engineer (Princeton: Princeton University Press, 
1954), on the battle over high-pressure steam and for- 
mation of combined bureau in 1939. NARA RG457, 
SRMN-084, 'The Evolution ofthe Navy's Cryptologic 
Organization," 3. A very useful survey is David Kahn, 
"Roosevelt, Magic and Ultra," Cryptologia 16 (1992): 

6. (U) Library of Congress, Papers of Stanford 
Caldwell Hooper: June 10, 1935, "McDowell, Contact 
scientists"; November 20, 1935, "Travel to Boston"; 
Box 18, "Johns Hopkins Atomic Energy," November 3, 
1937; and March 3,1933 to January 1, 1935, 
Correspondence with Redman, Jewett, "Contact scien- 
tists." NARA RG457, SRH-355, "Naval Security Group 
History to World War II," 268-269. 

%(U) NARA RG457, SRH-355, "Naval Security- 
Group History to World War II," 269, "Hooper to 
Wenger" November 1935, 270. January 2, 1936, "Bush 
Report." Office of Naval Research, Bush Comparator 
Patent file, #2,873,912. 

8.(U) A thorough search ofthe OP-20-G archives 
and the holdings ofthe NRL and, by implication, those 
ofthe Bureau of Ships, did not lead to the recovery of 
a copy of the original outline or Bush's first sketches of 
the Comparator. tSi, However, citations to the Bush 
documents were located on the old catalog cards ofthe 
NSA Technical Library. NSA, CCH, Card Catalog for 
Technical Library; Correspondence File: Bush 
Comparator 1936-38; Early Comparator Design 
Proposals: Bush Comparator 1936-37; Machine 
Proposals and Miscellaneous Material: Bush 
Comparator, 1936-1938; Navy Correspondence File 
and Historical Summary Bush Comparator, 1936-45; 
Six Project Reports: Bush Comparator, 1937-8; Bush 
Comparator Drawing #1-87 and Index, 1937; 
Summary of Materials on Symmetrical Sequences: 
Bush Comparator; Bush Comparator Drawings #88- 

153, 1937. f[3005]- (U) NSA CCH Series XII Z, 
"Memoranda on SIS, Formation of Cryptanalytic 
Group" from CCH Series XI K, Box 13, circa 1929- 


9.(U) NSA RAM File, January 28, 1936, DNC to 
Bureau of Ships, "Support Bush proposals," and July 
21, 1936, Bureau of Engineering toOP-20-G, "BuEng 
refuses Bush." NARA RG457, SRH-355, "Naval 
Security Group History to World War II," 269. 

10. (U) There is some indication the Bureau's men 
eventually outlined their solution, but no documents 
have been released. In the absence of any specifics, one 
can only guess at their alternative. Wenger remarked 
that the Bureau never really understood what he want- 
ed. NARA RG457, SRH-355, "Naval Security Group 
History to World War II," 269. 

11. (U) Susan M. Lujan, "Agnes Meyer Driscoll," 
Cryptologia 5 (1991): 47. James Rusbridger and Eric 
Nave, Betrayal at Pearl Harbor (New York: Summit 
Books, 1991). Cipher A. Deavours, and Louis Knih, 
Machine Cryptography and Modem Cryptanalysis 
(Dedham, Mass.: Artech House, 1985), 218. NARA 
RG457, SRH-355, "Naval Security Group History to 
World War II," 161, 247. The Holtwick M-i machine 
was in operation by mid-1937, perhaps earlier. 

12. (U) Jeff Wenger interview with W. S. 
MacDonald, March 1991. 

13. (U) Interviews with Waldron S. MacDonald. 
NARA RG457, SRH-355, "Naval Security Group 
History to World War II," 404. 

14. (U) Hagley Museum and library, Accession 
1825, Honeywell vSperry-Rand .Trial Records, 
Carton 67, William Radford, "Report on An 
Investigation ofthe Practicality of Developing a Rapid 
Computing Machine," October 15, 1939, Appendix III, 
List of Numbered References. Hagley Museum and 
Library, Accession 1825, Honeywell v Sperry-Rand , 
Trial Records, NCR Depositions, Bush to Deeds 
"Analyzing Equipment," May 19, 1938. Radford had 
been one ofthe "boys" at MIT who survived on the 
1930s version of "soft money" being a research assis- 
tant there from 1932 to 1939. Caldwell put him to work 
on the Rapid Arithmetical problem in early 1937. and 
he produced his report on "The Practicality of 
Developing a Rapid Calculating Machine," October 15, 
1939. Rockefeller Archives, Papers of Warren Weaver, 

Page 56 



January 16, 1946, letter, S.H. Caldwell to Weaver, 
"Center of Analysis," 4. 

15. (U) The machine did not have a name until 
some years later. Who borrowed the name 
Comparator from the nineteenth century MIT device 
remains unknown. 

16. (U) NARA RG457, SRH-355, "Naval Security 
Group History to World War II," 208, 247- 

17. (U) NSA RAM File, OP-20-G to Bureau of 
Engineering, "Plugboards for Reproducing Punch," 
July 7,1936. NSA, Lou Holland, "Development of 
Machine Processing in the Naval Security Group," 9. 
Holtvvick had created several small mechanical 
machines that cost less than $300 each for Japanese 
problems. The Bureau helped build them. NARA 
RG457, SRH-355, "Naval Security Group History to 
World War II," 210, 257, 261. 

18. (U) The patent claim was filed on April 22, 
1937. William F. Friedman and Vernon E. Whitman, 
Electric Control System for Tabulating Cards, 
Documents and the like, U.S. Patent 2,224,646, 
December 10, 1940. 

19. (U) NARA RG457, SRH-274, "Military 
Cryptanalysis." NARA SRH-004, "The Friedman 
Lectures on Cryptology." 

20. (U) David Kahn, Sewing the Enigma, (Boston: 
Houghton-Mifflin, 1991), 87, and Cipher A. Deavours, 
"The Black Chamber: A Column: LaMethode Des 
Baton," Cryptologia , 4 (1980): 240-247. There are 
reports that the Americans, including the Coast Guard 
group charged with attacking the ciphers of the rum- 
runners, were able to break into the simple commer- 
cial version of the Enigma during the 1920s. Malcolm 
F. Willoughby, Rum War at Sea, ( Washington: GPO, 

21. (U) David Kahn, Seizing the Enigma, (Boston: 
Houghton-Mifflin, 1991), 141. 

22. (U) W. S. MacDonald Interviews. 

23. (U) Letter to author from MIT registrar's 
office. In his first interview, MacDonald stated that he 
went to MIT in June 1937. But in his second interview 
he cited September as the month he began his duties. 
I have accepted the second date because it makes more 
sense in light of the previous pace of the work and 
MacDonald's desire to also enroll in the graduate pro- 
gram in the department. 

24. (U) Paul F. Ceruzzi, Reckoners: The Prehistory 
of the Digital Computer (Westport, Conn.: Greenwood 
Press, 1983)- 

25. (U) Michael R.Williams, A History of 
Computing Technology (Englewood Cliffs, N.J.: 
Prentice-Hall, 1985). Nancy Stern, From ENIAC to 
UNIVAC: An Appraisal of The Eckert-Mauchly 
Computers (Bedford, Mass.: Digital Press, 1981). 

26. (U) Because the Comparator read ten data 
columns at a time, more sophisticated IC tests could be 
done at the same rate as single letter tallying. Bush and 
Wenger may not have realized the potentials of the 
modified tabulators. The Comparator had five coun- 
ters as did tabulators. Thus, several fields could be 
processed in parallel. The navy may not have analyzed 
all the (n*(n-i)) combinations, but may have been sat- 
isfied with only a portion of the possible offsets of the 

27. (U) In some systems, messages were much, 
much longer. Thinking their Fish system was beyond 
attack, the Germans sent very lengthy reports on it. F. 
H.Hinsley etal., British Intelligence in the Second 
World War Vol. Ill, 1 (London: Her Majesty's 
Stationery Office, 1984), Appendix 2, 477. 

28. (U) Bush gave about one second for each line 
of printing. This time was estimated via the descrip- 
tion in the NSA RAM File, "MAC, Outlines # 17, 
70mm Comparator," and my knowledge of the 1938 

29. (U) Table 7.1 shows the power ofvarious 
devices, including the Comparator (run at various pro- 
jected speeds) relative to tabulators operating at their 
typical 120 comparisons a minute. A cell entry in the 
table gives the worth of the alternative machine in 
terms of the hypothetical raw power of the number of 
tabulators. Thus, a sorter running at full speed was 
worth three tabulators while the speed of a typical tele- 
type system of the era was five times greater than the 

30. (U) On tape speeds and densities, Hagley 
Museum and Library, Accession 2015, Remington- 
Rand, ERA materials, S. Ruebens, "Investigation of 
Solid Acoustic Delay Lines," Contract Nobs 28476, 
August 1, 1947, 1. The Colossus read at 5,000 charac- 
ters a second. Brian Randell (ed.), The Origins of 
Digital Computers: Selected Papers (New York: 


Page 57 


Springer-Verlag, 1982), 349. It is not certain that this 
means that 5,000 serial characters passed the reading 
head of its tape scanner. The Robinsons, the British 
versions of the tape-optical machines (but ones for 
binary comparisons), read at 2,000 a second. Tape 
readers used by the navy in WWII ran at about ten 
characters a second. The 1948 figure is in Samuel S. 
Snyder, "Abner: The ASA Computer, Part I: Design," NSA 
Technical Journal 25 (1980), 59, 

31. (U) Avery important spinoff of the navy 
Comparator was a 1938 project at NCR. Desch used 
35mm film with punched holes on an optical compar- 
ing device for a utilities billing machine. Hagley 
Museum and library, Accession 1825, Honeywell v 
Sperry-Rand Trial Records, August 16, 1938, Desch to 
Williams, "Laboratory work" and August 30, 1939, 
Desch to Williams "Work at Laboratory." 

32. (U) Bernard Williams, "Computing With 
Electricity, 1935-1945," (Ph.D. Thesis, University of 
Kansas, 1984). Final OSRD Report, Div. 17, George E. 
Beggs Jr. and F. L. Yust, "Development and 
Application of Electronic Counting Circuits," 1946, 
especially Chapter 9. 

33. (U) Several late 1930s projects at MIT explored 
magnetic memory and many variations of storage 
based on electrical charges. See Hagley Museum and 
Library, Accession 1825, Honeywell v Sperry-Rand 
Trial Records, Carton 67, William Radford "Report on 
An Investigation of the Practicality of Developing a 
Rapid Computing Machine, October 15, 1939." 

34. (U) William Aspray (ed.), Computers Before 
Computing (Ames, la.: Iowa State University Press, 

35. (U) Michael K. Buckland, "Emanuel Goldberg, 
Electronic Document Retrieval, and Vannevar Bush's 
Memex," JASIS 43 (1992): 284. On the 1937 and 1940 
patents and their history, Hagley Museum and 
Library, Accession 2015, unprocessed ERA materials 
from Sperry Archive, November 1, 1949, Memo to File, 
Selector Infringement Search, and Accession 1825, 
Honeywell v Sperry-Rand Trial Records, August 13, 
1937, Bush to Deeds, and October 25, 1937, Research 
Corporation to Deeds. 

36. (U) NSA RAM File, OP-20-G to Bureau of 
Engineering, "Rapid Equipment," March 29, 1938. 

37. (U) A density often or eleven per inch was 
assumed as indicated by the description of the later 
army version of 1944 which reached twelve and one- 
half per inch with an average over the entire tape of six 
and one-quarter per inch. On the cost of the Eastman 
tape, NARA RG457, SRH-355, "Naval Security Group 
History to World War II," 276, and NSA RAM File, 
CNO to Bureau of Engineering, April 29, 1938. 

38. (U) In later models of the Comparator, the 
stepping could be from one to ten characters after each 
pass. NSA RAM File, "MAC. Outlines #17, 70mm 
Comparator," April, 1947. 

39. (U) Howard Aiken's early computer, the ASCC, 
used a tape rig similar to the one chosen for the 
Comparator. During World War II the Americans and 
the British also used a pulley and loop system for the 
follow-ons to the Comparators, indicating that it was a 
sound method. Michael R. Williams, A History of 
Computing Technology (Englewood Cliffs, N.J.: 
Prentice-Hall, 1985), 245. Brian Randell, "Colossus: 
Grandfather of the Computer," in B. Randell (ed.), The 
Origins of Digital Computers (New York: Springer- 
Verlag, 1982), 350. 

40. (U) Interviews with W. S. MacDonald. 

41. (U) Paul F. Ceruzzi, Reckoners: The Prehistory 
of the Digital Computer (Westport, Conn.: Greenwood 
Press, 1983). Charles S. Bashe, etal, IBM's Early 
Computers (Cambridge, Mass.: MIT Press, 1985), 36- 


42. (U) Waldron MacDonald claimed he had to 
rework all the circuits and that binary switching 
speeds did not exceed 5,000 per second. MacDonald 
interviews, 1987-1991. 

43. (U) The navy's later Letterwriters and its 
Copperheads used a modified baudot coding. NSA 
RAM File: "Machine Comparisons," June 1946; 
Communications Intelligence Technical Paper 42, 
"Copperhead I Theory and Copperhead I Equipment"; 
and Communications Intelligence Paper 41, 
"Copperhead I Punch and Copperhead I Scanner." 

44. (U) NSA RAM File, July 18, 1938, Safford to 
Bush, "Machinery Arrived." 

45. (U) NSA RAM File, Wenger Report, "Bush 
Visit" April 25, 1938. Waldron S. MacDonald inter- 
views. NARA RG457, SRH-355, "Naval Security Group 
History to World War II," 299. 

Page 58 



46. (U) Interviews with Waldron S. MacDonald, 


47. (U) Library of Congress, Papers ofVannevar 
Bush, Box 67, MacDonald to Bush, July 25, 1939. And 
interviews with MacDonald. 

48. (U) Office of Research and Inventions patent 
application sheet of 10-29-46 lists April 1938 as the 
time of the first successful witnessed run. ONR, patent 
file for Bush Comparator, #2,873,912. 

49. (U) NARA RG457> SRH-355, "Naval Security- 
Group History to World War II," 272, 276. NSA RAM 
File, OF-20-G to Bureau of Engineering. "Rapid 
Equipment," April 29, 1938, and May 17, 1938, 
"Comparator Equipment." 

50. (U) NARA RG457, SRH-151, "Military Study: 
Communication Intelligence Research Activities," 
022, indicates that the first Bush machine was paid for 
out of a special fiscal 1938 allocation. The proposed 
budget for 1939 contained a request for a similar 
amount, but it may have been for a "payback" for the 
first expenditure. 

51. (U) NR4. RAM File, September 16, 1938, OP- 
20-G Bureau of Ships, "Budget Request" 

52. (U) The exact amounts spent and budgeted for 
the Comparator in 1938 and 1939 remain unknown. 
NARA RG457, SRH-355, "Naval Security Group 
History to World War II," 276, 240. The budget 
request for fiscal 1939 included items for building new 
machines and new components such as a rapid "loca- 
tor." This may have been a means of locating code 
groups, a function later embodied in the machines 
designed by Lawrence Steinhardt, or it may have been 
an automation of the methods of overlay sheets to 
determine code-wheel orders and setting as used by 
Mrs. Driscoll. RAM File, CNO to Bureau of 
Engineering, September 16, 1938, "Development of 
Special Communications Devices." 

53. (U) The machine arrived at the Navy Yard on 
June 24. It had been badly jostled on the trip and 
refused to run. NSA RAM File, July 18, 1938, OP-20-G 
to Bush "Machine Has Arrived." It took Dulong some 
three weeks to tease the machine into its first non-MIT 
test run. NRL, Bush Comparator patent application 
file, October 29, 1946, 1. 

54. (U) NSA RAM File, July 18, 1938, Safford to 
Bush, "Machinery Arrived." NRL patent application 
file, October 29, 1946, 1. 

55. (U) NSA RAM File, May 17, 1938, "Comparator 
Equipment," and July 18, 1938, Safford to Bush, 
"Machinery Arrived." 

56. (U) Jack Sweetman, American Naval History 
(Annapolis: Naval Institute Press, 1984), 156. 

57. (U) Jeffery M. Dorwart, Conflict of Duty: The 
United States Navy's Intelligence Dilemma 1919-1945 
(Annapolis: Naval Institute Press, 1983), 93, 99. NSA 
RAM File, OP-20-G to Bureau of Engineering, March 
22, 1938 and June 24, 1938, "IBM Purchases." Cipher 

A. Deavours, and Louis Kruh, Machine Cryptography 
and Modern Cryptanalysis (Dedham, Mass.: Artech 
House, 1985), 212. NSA, Theodore M. Hannah, "Frank 

B. Rowiett; A Personal Profile," 116. Rear Admiral 
Edwin T. Layton, And I Was There: Pearl Harbor and 
Midway - Breaking the Secrets (New York: William 
Morrow, 1985). NARA RG457, SRMD 019, "The 
Panay Incident." 

58. (U) NSA RAM File, May 17, 1938, "Comparator 
Equipment," and CNO to Bureau of Engineering, 
September 16, 1938, "Development of Special 
Communications Devices." 

59. (U) Ironically, MIT would soon provide the 
army's cryptanalytic group with an electrical engineer, 
Leo Rosen, who quickly became the leader of the 
group that constructed a model of Purple and that 
became the electronics research group in the SIS. NSA, 
Theodore M.Hannah, "Frank B. Rowiett: A Personal 
Profile," 18. 

60. (U) NSA RAM FUe, Safford to Bush, December 
10, 1938, "Fine Job on Comparator." 

61. (U) NARA RG457, SRH-355. "Naval Security 
Group History to World War II," 300. 

62. (U) NSA RAM File, OP-20-G, "List of statisti- 
cal machinery," December 1, 1939. 

63. (U) Library of Congress, Papers ofVannevar 
Bush, Box 67, MacDonald File, Bush to MacDonald 
"OP-20-G Project," August 31, 1940. NARA RG457, 
SRH-355, "Naval Security Group History to World 
War II," 270, 405. 

64. (U) Daniel J. Kevies , The Physicists: The 
History of the Scientific Community in Modern 
America (New York: Knopf, 1978), 296. 


Page 59 


65. (U) Ronald W. Clark, Timrd (Cambridge: MIT 
Press, 1965). 

66. (U) Charles S.Bashe, etal., IBM's Early 
Computers (Cambridge, Mass.: MIT Press, 1985), 
Chapter 2. 

67. (U) Hagley Museum and Library, Accession 
1825, Honeywell v Sperry-Rand , Trial Records April 
16, 1942, "Conference on fire-control projects." The 
ENIAC used a "memory" system from RCA that was 
quite like what was used on some OP-20-G and SIS 
machines and a counting circuit invented by one of 
RCA's men, Igor Grosdoff. N. Metropolis, et al., (ed.), 
A History of Computing in the Twentieth Century 
(New York, 1980), 467. Hagley Museum and Library, 
Accession 1825, Honeywell v Sperry-Rand , Trial 
Records, May 19, 1938, September 18, 1944. "Army 
seizes Grosdoff patent." 

68. (U) Bernard Williams, "Computing With 
Electricity, 1935-1945," (Ph.D. Thesis, University of 
Kansas, 1984), 317. Hagley Museum and Library, 
Accession 1825, Honeywell vSpeiry-Rand Trial 
Records, June 6, 1942, Weaver to participants "Cancel 
fire control work." 

69. (U) The fear of defeat was quite real. Bletchley 
Park had scores of buses at the ready to race its staff to 
port cities where they were to board fast ocean liners 
for the United States and Canada. 

70. (U) Laurance F. Safford, "Rhapsody in Purple," 
byDundas P.Tucker, Cryptologia 6(1981): 196,220. 
NSA Theodore M.Hannah, "Frank B.Rowlett: A 
Personal Profile," 18. NSA RAM File, Part H of Report 
to J. N. Wenger, Capt. USN, "Resume of the Dayton, 
Ohio Activity During World War II," (presumably a 
continuation of the Meade Report), 1. In contrast to 
Safford's interpretation and on America's first contact 
with British Enigma achievements, NARA RG457, 
SRH-361, "History of the Signal Security Agency, 
Volume Two, The General Cryptanalytic Problems," 
019-021, 272. NARA RG457, SRH-145, "Report of the 
Technical Mission to England," April 11, 1941, 002- 
004. But on how little Enigma ability OP-20-G had as 
late as the summer of 1943, NARA RG457, SRH-403, 
"Selections from the Cryptologic Papers of Rear 
Admiral J. N. Wenger, USN," 072-3. 

71. (U) Rear Admiral Edwin T. Layton, And I Was 
There: Pearl Harbor and Midway -Breaking the 
Secrets (New York: William Morrow, 1985), 78. 

72. (U) NARA RG227, OSRD, Office of the 
Chairman, Box 1, Bush to Safford, October 28, 1940, 
"Project Agreement." NSA RAM File, Safford to Radio 
Division, Bureau of Ships, November 2,1940, "Bush 
Project." NARA RG457, SRH-355, "Naval Security 
Group History to World War 11," 404. 

73- (U) NSA RAM File, January 2, 1942, Safford to 
Howard, "Comparator Received 12-24," and January 
6, 1942, BuShips to CNO, "Manual for Comparator." 

74. (U) Again, this may also have been a means of 
automating the use of the sheets used to attack wheel 

75. (U) The punches were redesigned and reman- 
ufactured several times during the war. After MIT 
made a try in 1940-41, the Gray manufacturing com- 
pany made a version; then NCR redid them; then new 
designs were drawn for the postwar era. NSA RAM 
File, Communications Intelligence Technical Paper- 
42, "Copperhead I Theory and Copperhead I 
Equipment." NSA RAM File, 0P-20-G to NCML-NCR, 
February, 1945. 

76. (U) NARA RG457, SRH-355, "Naval Security 
Group History to World War II," 430. On Steinhardt 
visit, Smithsonian Institution, History ofComputers 
Project, Interview with Howard Campaigne, 19. 

77. (U) Some $25,000 was promised to NCR in 
November 1940. Some of it may have come from the 
original NDRC grant, but Bush apparently secured an 
additional grant for some thirty copies of the counters 
although the NDRC was not to become involved in 
production. NARA Suitland, OSRD Contract Files, 
OEM 275 November 28, 1941, "NCR-MIT counters." 
Hagley Museum and Library, Accession 1825, 
Honeywell v Sperry-Rand , Trial Records, December 
2, 1941, NDRC D3 to Desch-NCR, "30 counter print- 

78. (U) Hagley Museum and Library, Accession 
1825, Honeywell v Sperry-Rand , Trial Records, 
January 29, 1942, Desch to NCR "Secret Work," and 
February 21, 1942, Desch, "Report to Management." 

79. (U) Gordon Welchman, The Hut Six Story: 
Breaking the Enigma Codes (New York: McGraw-Hill, 
1982). David Kahn, Seizing the Enigma (Boston: 

Page 60 



Houghton-Mifflin, 1991). The role of captures in the 
long-awaited break into Japanese army systems is dis- 
cussed in the very impressive Edward J. Drea, 
MacArthur's Ultra (University of Kansas Press, 1992). 

80. (U) On the state of American readiness and 
some hints about the roles of Mrs. Driscoll and the 
team at SIS, compare James Rusbridger and Eric 
Nave, Betrayal at Pearl Harbor (New York: Summit 
Books, 1991), and the more scholarly Edward J. Drea, 
MacArthur's Ultra (University of Kansas Press, 1992). 
NARA RG457, SRH-355, "Naval Security Group 
History to World War II," 440, 442. NSA RAM File, 
OP-20-G toOP-20-A, "Meeting with Dr. Howard," 
November 5, 1941. However, Howard was told that the 
Driscoll problem was not of high priority. NSA RAM 
File, BuShips to Howard, November 11, 1941. 

81. (U) NSA RAM File, November 3 and 5, 1941, 
Howard OP-20-G Reports on Meetings, and 
November 14, 1941, Bureau of Ships to Howard. 

82. (TS//SI) MAC. Outlines #1, April 1947, 
Electroanagrammer ("Gee Whizzer"), stored inCCH 
Series XII Z. Leroy Wheatley, "Brief Description of 
Analytic Equipment, Fourth Installment," 20 
September 1954, "NC MACHINES," stored inCCH 
Series XII Z. 

83. (TS//SI//REL) Samuel S.Snyder, "Famous 
First Facts: Part I: Pre-Computer Machine 
Cryptanalysis," unpublished, stored in CCH Series XII 
Z. "Gee Whizzer," circa 1947, stored in CCH Series XII 

84. (U) S.Snyder Interview with Lou Holland, 
circa February 2, 1972, stored in CCH XI K, Box #10, 
Famous First Fact folder. Wheatley, "Brief Description 
of Analytic Machines, Fourth Edition," 20 September 
1954, "NC Machines." 

85. (U) NSA S 337/348 "Additive Machines- 
Historical Summary Of," Stored in CCH Series XII Z. 
The first M4was operational at the end of 1941. It 
worked well for a time, but had to be replaced. 

86. (U) Photographic plates for data entry had 
been used on the Cinema Integraph and were being 
explored for use in the Analyser and the electronic 
computer at the Institute. NSA RAM File, January 5, 
1942, Howard toOP-20-G, "Report: Glass Plates." 
Arnold Dumey letters to Brian Randell, 1975, "defor- 

87. (U) NARA RG457: SRH-279, "0P-20-G File 
Communications Intelligence Organization, 1942-46:" 
SRMN-084, "The Evolution of the Navy's Cryptologic 
Organization," 2; and SRH-306 "OP-20-G Exploits 
and Commendations World War II." 

88. (U) Interview with Jeff Wenger, 1991. 

89. (U) NARA RG457, SRMN-083, "Military 
Study of Secret Radio Calls, January 1938." 

90. (U) All other major nations used the same 
techniques and had done so since World War I, but 
Wenger was America's advocate. NARA RG457, 
SRMN-083, "Military Study of Secret Radio Calls, 
January 1938." 

91. (U) NARA RG457, SRH-083, "Military Study 
of Secret Radio Calls, January 8,1938 by J. N. 

92. (U) On automatic direction finding equipment. 
NARA SRH-197, "US Navy Communications 
Intelligence, Organization, Liaison and Collaboration 
1941-1945," 38. 

93. (U) Some reports indicate that in late 1941 the 
vast majority of the "G" workforce was busy with T/A 
rather than cryptanalysis. NARA RG457, SRH-403, 
"Selections from the Cryptologic Papers of Rear 
Admiral J. N. Wenger, USN." 

94. (U) NARA RG457, SRH-279 "OP-20-G, 
Communication Intelligence Organization 1942- 

95- (U) NARA RG457, SRH-305, "The Undeclared 
War: The History ofRI," 15 November 1943, by 
Laurance F. Safford, Captain, U. S. Navy. 

96. (U) NARA RG457, SRH-403, "Selections from 
the Cryptologic Papers of Rear Admiral J. N. Wenger, 
USN," 60. BuShips, NXS329 (945) 5-6-43, Pulse 
Spotter Equipment, Philco Corporation. The remain- 
der of the war. Library of Congress, Papers of 
Vannevar Bush, Box 52, Howard to Killian, March 21, 
1946, "rapid selector." 


Page 61 

This page intentionally left blank 

Page 62 


Chapter 3 
(U) Bush's Dream Does Not Come True 

(U) A Look Ahead to Peace 

(U) America's entry into the war led to the 
release of torrents of money for codebreaking. A 
surprisingly large allocation was given to the 
Bureau of Ships for cryptanalytic machine devel- 
opment. Unfortunately, "G" and the bureau were 
not prepared, and they were unable to immedi- 
ately establish a well-coordinated project that 
could compensate for the years of lost opportuni- 
ties. The funds came too late to set up a long-term 
development program, and "G's" resources had to 
be devoted to cryptanalytic fire-fighting. Until 
very late in the war, OP-20-G's computer activi- 
ties were driven by emergencies. Until late 1943, 
"G's" brilliant mathematicians and engineers did 
not even have the time to think of machines that 
went beyond Bush's mid-i930s ideas or to plan 
their move from analog to electronic digital tech- 

(U) After they made their great electro- 
mechanical contribution to the Ultra problem, 
the Bombe, they had more time. They built a few 
path-breaking digital electronic machines, and 
they began to lay plans for a long-term computer 
program. However, by the end of the war they 
had not been able to turn Bush's faith in micro- 
film into advanced and reliable machines. 

~t$f One important consequence of the 
absence of the ultra-fast digital machines was that 
Hooper's dream of relying upon pure statistical 
and mathematical cryptanalytical techniques had 
to be deferred. Like England, the United States 
had to rely upon the most expedient cryptanalyt- 
ic as well as technological solutions. During 1942 
and 1943, OP-20-G's "M" group was unable to 
become a think-tank for pure methods. They and 
the cryptanalysts had to depend upon "what 
worked." And what worked in cryptanalysis were 

brute force techniques that called upon massive 
data processing and hunch playing. At the same 
time, the engineers at "G" and the Bureau of Ships 
had to choose the easiest hardware solutions to 
meet the demands of the hard-pressed code- 

(U) A series of increasingly complex and pow- 
erful devices began to emerge in early 1944. But it 
was not until 1945 that OP-20-G's young engi- 
neers could begin to think of creating a multipur- 
pose Rapid Machine, and when "M" could begin 
the exploration of the frontiers of mathematical 
cryptology. There were more than technical rea- 
sons for the gap between what Wenger and 
Hooper wanted and what was achieved during 
the war. OP-20-G's World War II machine effort 
began in crisis, had to respond to immediate 
cryptanalytic needs, and continued to be driven 
by rapidly shifting demands. 

(U) January 1942: Too Much Too Late 

(U) Soon after Pearl Harbor, the bureau 
gained the funds to support all the ideas that had 
been put forward by the various groups in OP-20- 
G and at the Navy Yard. Contracts were let to IBM 
for more tabulator equipment and for the cre- 
ation of a host of new special attachments. Those 
at the bureau and OP-20-G who favored electro- 
mechanical equipment received recognition 
when IBM was also awarded a very large contract 
to develop a set of new machines to automate the 
processing of incoming data. At the same time, a 
group of navy engineers in Washington was 
allowed to build some electromechanical analysis 
machines of their own design. 

(U) The bureau had enough resources to pre- 
vent Bush's Comparator from being locked away 
to die in the secret workroom at MIT. The bureau 


Page 63 


made an arrangement with Eastman-Kodak to 
work on all the microfilm and plate ideas. This 
meant that Howard and his group were to be 
helpers not supervisors for the teams Eastman 
hastily put to turning ideas into machines. 1 When 
Joseph Wenger returned and established the "M" 
group under Howard Engstrom, he tried to regain 
control over the automatic machines. But it took 
some time to organize "M." In fact, if it had not 
been for the crisis in the Atlantic and the attitude 
of Ralph Meader, the man the bureau had 
assigned to supervise the machine contracts, 
Wenger would not have been able to reassert OP- 
20-G's power over machine development. 

(U) In the first months of 1942, Meader ran a 
freewheeling one-man operation for the bureau. 
Despite his freedom, he began to experience the 
frustrations that had led Hooper and Bush to try 
to throw off the heavy hands of the navy's bureau- 
cracy in the 1930s. He came to feel that the com- 
panies were unresponsive, and he compiled a list 
of complaints. 

certainly not a machine that was leading, as were 
the Bush designs, to the use of digital circuits. 

(U) The machine was electrical, not electron- 
ic Eastman's team realized that a pulse-based 
system, even with the plates, would be too com- 
plex. Thus, an electric measuring system was 
built into the IC machine. There was no counting, 
just a recognition that enough light had penetrat- 
ed to the photocell. The analyst would then tally 
the overlapped dots or find their locations within 
a message. 4 

(U) Although comparatively simple, the IC 
machines had deficiencies. It was very difficult to 
coax the data camera to place the dots on the 
plates in perfect alignment; that problem contin- 
ued throughout the war. The IC machines them- 
selves had to be redesigned and reworked during 
1943. Perhaps as many as one half of all the 
machines were inoperable at any one moment. 5 

(U) Haste and Conjiision 

(U) A Giant Step Backwards 

(U) When the bureau went to Eastman in 
early 1942, no one had expected frightening 
delays or a need for radical redefinitions of "G's" 
machines. Eastman's technical and managerial 
reputation pointed to a speedy solution to the 
problems that had halted the work at the 
Institute. Thus, it was natural for the bureau to 
turn to it when Howard seemed to admit that he 
could not solve the problems of the proposed 
microfilm Comparator. But Eastman would not 
meet the bureau's expectations. 

(U) With help from John Howard's men, 
Eastman was able to ship the first version of what 
became known as the Index of Coincidence 
Machine before 1943. 2 The IC Machine was a rel- 
atively simple plate-based device that looked 
more like the early 1930s astronomers' instru- 
ments than Bush's Comparator. 3 The IC Machine 
did its job, but it was not automatic, and it was 

(U) Eastman's work began in haste, was not 
well supervised, and, as a result, was not ade- 
quately documented. As a consequence, even the 
military services were confused about the names 
of the machines Eastman proposed or delivered 
before 1943. 6 Some of the first lashed-together 
models are only vaguely remembered. But the 
documents that remain reflect the desperation to 
produce machines. 

(U) One of those was aversion of a Bush IC 
machine constructed during 1942. When it ran, it 
shook the entire laboratory; perhaps that is why it 
never appeared in OP-20-G's machine center. 
The device did not use the ideas for tape drives 
that Bush and Howard had explored. Rather, two 
large message tapes were wound around a large 
hydraulically controlled drum. As the drum spun 
at a very high speed, the tapes inched back and 
offset themselves. When the photoelectric detec- 
tor sensed a "hit," the drum slammed to a stop. 7 

Page 64 



-i$i Some other alternative concepts for 
machines came from the groups at Eastman as 
they searched for ways to make Howard's original 
suggestions turn into hardware. Trying to help 
the navy in its attack against the very stubborn 
Japanese additive code systems, the Eastman 
engineers drew up initial plans for a combination 
of tape readers, electronic circuits to strip the 
additives, and a set of whirling disks that were to 
hold frequently used code groups. An Edgerton 
flash lamp was to help to see if the stripped text 
groups matched one of the codes on the disks. 8 

(S//SI) The disk contraption was not deliv- 
ered to the navy, but Eastman's initial attempt to 
turn Bush's Rapid Document "Selector" 
(Information Machine) ideas into a useful mili- 
tary machine arrived in Washington before the 
end of 1942. Tessie was Eastman's first great con- 
tribution. Tessie began as an attempt to finally 
turn the architecture and the fundamental tech- 
nologies of the microfilm Selector into a machine 
that worked. But Tessie became another prag- 
matic compromise. Bush's ideas were too difficult 
to put into practice. They could not be changed 
into hardware quickly enough to meet the press- 
ing cryptanalytic needs of the first years of the 

(U) Tessie Wouldn't Either 

■fS}"John Howard made sure that Eastman 
knew what he wanted well before the formal con- 
tracts with the Bureau of Ships were signed. He 
told the men in Rochester that in order to deliver 
something of value, in what was hoped was a rea- 
sonable time, the new military "Selector" was to 
be a special purpose device to perform an impor- 
tant but minimally challenging cryptanalytic 
function. Its job was to locate four-character code 
groups (tetragraphs), not to count them. Finding 
and giving the location of groups was a "quick and 
dirty" version of IC analysis. "Locations" pointed 
to the possibility that two messages were in 
"depth." 9 

(S) T essie's logic and architecture were like 
those of a Selector rather than a Comparator. It 
used two 35-millimeter films. One sped past an 
optical reading station while the other remained 
stationary and acted as an identification mask. 
After the fast film made a complete revolution, 
the mask film was stepped one frame. If a desired 
tetragraph was located, a signal was emitted. 
Then, a strobe circuit signaled a high-power 
Edgerton flash to send light through the identifi- 
cation portions of the two data films. The light 
would register the location of a matched group. 
After the run was completed, the new film was to 
be quickly developed, then sent to the analysts 
who would trace the groups and begin their 
attempt to break the codes. 10 There was no abili- 
ty to reproduce the code groups, and there was no 
ability totally. Those features, Howard knew, 
would ask too much ofthe engineers working 
under pressure - even of Eastman's experts. 11 

(U) The Eastman group put in a great deal of 
overtime and was able to send a machine to 
Washington in September 1942. RAM-2, or 
Tessie I, was a large and ungainly thing that was 
more than six feet long and almost as high. On 
one end was the drive mechanism for the data 
microfilms. On the other was a huge round canis- 
ter-like component that housed some ofthe elec- 
tronics and the photographic reproduction equip- 
ment. On top ofthe canister was a rack of tubes 
that could not be squeezed into the machine's 
frame. 12 

-^Although ugly, Tessie raised expectations 
about the Eastman portion ofthe RAM program. 
A great deal of equipment was ordered to support 
Tessie's work. Fifty- and sixty-foot metal develop- 
ing trays, film drying racks, and hundreds of 
pounds of chemicals began arriving at the 
cramped "G" headquarters. 13 

-(&)- Tessie failed, however. In its first runs it 
missed almost all the coincident tetragraphs. 
When it did find a "hit," it refused to produce the 
record of it on its internal film. The Edgerton- 


Page 65 


1943 to replace many of the 
circuits and to design a new 
flash system. The machine 
did begin to do a bit of work. 
It was put to use in an attempt 
by the Americans to attack 
the Enigma by searching for 
tetragraphic repeats. 16 Tessie 
proved of some use in the 
next few months although it 
continued to misbehave. 
Because of continued prob- 
lems, it was replaced as soon 
as possible and changed into 
an even greater analog retro- 



type flash system would not function. Even the 
Washington, D.C., water supply refused to coop- 
erate in the film development process of the two 
"data" films. Chemical imbalances in the water 
were making the tiny dots the Tessie light bank 
data camera produced spill over onto each other, 
making recognition impossible. That added to the 
difficulties of making the camera light bank 
behave. Some of the problems with the camera 
were fixed, but it took longer to find a way to com- 
pensate for the chemistry of the District's water 
supply. 14 

-f T8 > One of Tessie's weaknesses was very 
embarrassing for the engineers. It missed "hits" 
that were too close together. The special warning 
circuit Howard and his men had devised to solve 
the problem would not behave. 15 

(SJThroughout fall 1942 Tessie was too unre- 
liable to be used as an operational machine. 
Finally, it was decided to make a major invest- 
ment in its repair. It took Larry Steinhardt and 
his crew in Washington almost all of January 

(U) Tessie's New Hat 

€¥S}-The navy's Tessie 
was turned into a machine to 
perform a very simple type of 
search for "isomorphs." Asa result, the new 
Tessie used only a few ideas from Bush's 1930s 
proposal for a navy "Symmetrical Sequence" 
engine. In its new life, it no longer reproduced hit 
locations on film. The flash camera was aban- 
doned in favor of a punch. 18 

(TS//8I) Codebreakers search for anything 
that is nonrandom. When they can find patterns 
that are obviously not produced by chance alone, 
they have at least a beginning of an attack on an 
enemy system. Repeats of phrases or even words 
are one of the signals that additional analysis 
might lead to a successful understanding of an 
enemy's cipher. One of the ways cryptanalysts 
located repeated messages, or repeated groups 
within messages, was to search for what the army 
called "isomorphs" and the navy called "symmet- 
ric sequences 

» 19 

CCS//SI) Because cipher machines or additive 
systems are designed to hide repeats of words, 
looking for exact matches in small amounts of 
text is usually a waste of time. But what may be 

Page 66 



found are patterns. If the word "BATTLE" is 
treated as a sequence of letters and recorded as 
"ABCCDE," an analyst may search for a repeat of 
the pattern. The pattern may appear despite the 
ability of additives to disguise the underlying 
code group. 

(TS//SI) Isomorphic attacks are expensive 
"long-shots." They were time-consuming because 
of the need to recode message texts, and identify- 
ing an isomorph only led to a probability that a 
depth had been found. But the method had 
proven of value to the American cryptanalysts 
since at least the 1930s. 

-£FS> When the first Eastman Tessie was out- 
classed as a locator of exact match, four-character 
groups by its replacement, Icky, OP-20-G decided 
to have Tessie turned into an emergency 
Symmetric Sequence Machine. What emerged 
from the workshops in early 1944 became known 
as Tessie SS. The reborn Tessie was still a six-by- 
seven-foot monster, but it had been stripped of 
many of its most sophisticated components. 
However, it finally worked, and it saved a great 
deal of analyst and, as important, recoding 
time. 20 

mechanism like the ones built for the paper tape 
Comparator. When the photocell spotted a 
repeated letter, it ordered the two tapes to be 
punched with tiny holes. When the entire six- 
character-per-second run through the message 
was completed, the paper tapes were removed, 
placed on a viewer, and searched for patterns of 
red dots that would indicate where an isomorph 
had been found. Tessie SS was much less elegant 
than the original, but it functioned successfully. 

(U) You Can Use Some ojthe Technology 
Some of the Time, But... 

(U) Soon after the original Tessie was deliv- 
ered to Washington in fall 1942, it was realized 
that it would never be a success. A radical 
redesign would be necessary. Avery different 
design was needed if the navy was to have a suc- 
cessful high-speed microfilm machine. 

(U) Eastman assigned a new crew to the task 
in early 1943, but the company was unable to 
deliver a machine, Icky, until October 1943. Then 
it took another few months for the men in 
Rochester to develop an efficient camera to pro- 
duce reliable microfilms for the new Icky. 

(T9//3I) The new Tessie had a small reel for 
the 35mm message film and one for a mask that 
contained patterns for the letters A to Z. The pho- 
toelectric scanning system's first task was to iden- 
tify the first character of text on the film. Then, it 
scanned twenty characters of the message as the 
alphabetic mask film sped by. If it found a repeat 
of the first of the twenty characters, it signaled 
that the other half of the machine should get to 

(TS//SI) At the other end of Tessie SSwas 
that original huge round canister. But now it con- 
tained a roll of plain 70mm paper tape and a roll 
of the black-red tape that Bush had used on his 
paper tape Comparator, not unexposed film. 
Instead of the strobe system for reproducing the 
location codes for a hit, there was a punching 

-£E8tAsthey were designing abetter camera 
system, the Eastman group received some more 
depressing news. After its first rounds of tests in 
Washington, Icky needed to be reworked. It had 
to be shipped back to Rochester because of the 
great amount of repair and redesign that was 
needed. 21 On its return to Washington, the 
machine became an important tool for OP-20-G; 
but its success depended on its being, in some 
ways, another retrogression. 

(TS// S I) First, to allow it to be of immediate 
and reliable use, it had been designed to be much 
more limited than the Comparator or even 
Tessie. 512 Icky was a film version of the IC plate 
machine, and it was able to do a primitive analog 
ICtest, but it just located. It did not make copies 
or count. It had no reproducing cameras and it 


Page 67 


had no counter-printer. When enough 
light was registered, the machine lit a 
signal lamp and stopped. Its operator 
used a hand crank to turn the films 
back towards the identified coinci- 
dence. At the point where "enough" 
light came through the two films 
another signal lamp was lit. Then the 
operator used a screen to read the loca- 
tion marks printed on the margins of 
the films. 

(U) From its beginnings as a bench 
model, Icky evolved into a chest-high 
box as wide as a refrigerator. On its top 
was a screen to view the located mes- 
sages. Next to the screen were the reels 
and rollers for two 35mm microfilms. 
Underneath were the mechanisms that 
sped one of the films past the other 
and the ratchets that stepped the index 
film after each pass of the fast tape. 
Icky's optical sensing gate was 
designed to allow the location of mes- 
sage patterns of up to thirty columns of 
data. Typically, a bright light was 
pointed through masks and lenses 
which segregated the light into thirty 
parallel columns. Iflight penetrated 
films, it was directed to thirty small 
which then sent the light beam to their photocells. 

(U) The light management portion of the 
machine was complex and demanded perfect 
alignments. It was the demands of that photocell 
system that led to Icky's having only forty 
columns of data per inch of film, a density far less 
than Bush had promised the navy. 

(U) More significant, Icky was not a digital 
machine. Like OP-20-G's other World War II 
microfilm devices, it wandered back to the use of 
analog circuits. But it did have a plugboard and 
resistor matrix system that allowed the selection 
of many different combinations of coincidences. 
Polygraphs of long lengths, or patterns ofidenti- 



..Ph. • 


mum. wdek * 

MOTOH $w>to 



;," SWITCH ~ 

E * - «. 

*- fa wvn ' 






SPEEG jr ' 

|K¥£BSe - "" 

*"■ '■■■;, ' ■■'"■' '.-■' ■'-'" '■■" "*% 

&^\ ■ , ' "' «■'■'■-'■ y'**^ 



■ TCH 

■■ ^;-^t}*jvf*~^ : '. 

*% . *t, . 

. '4 ~'-S$SP*BT' 


the two 

cal subgroups, or single-letter coincidences could 
be identified. 

(U) Icky had another feature that went 
beyond the original Selector. Its coding system 
could be changed and its circuits switched so that 
it responded to the absence of light rather than its 
presence. 23 The navy's men found the blackout 
method much more efficient when the job was to 
search for coincident areas (such as code groups) 
rather than single columns of data. With its use, 
they could pack more than one letter in a column. 
They could register a five-letter (or number) mes- 
sage group. 

(U) In the blackout system, the two tapes were 
reciprocally coded so that a matched column 
would admit no light. Atwo-of-five character 

Page 68 



code allowed the use of that reciprocal scheme, 
but Ick/s scanner could also accommodate 
Bush's older one-of-twenty-six pattern. 24 

(U) A Machine for Mrs. Driscoll's Special 

(U) Eastman designed and built another of 
the very few types of microfilm machines used by 
OP-20-G during the war. The Hypo assignment 
came in a rush, and, like Tessie, it took a year to 
complete. 25 The first Hypo was not in operation 
until October 1943. 26 

(U) The name, Hypo, came from the 
"Hypothetical Machine" proposal drawn up in 
response to the early requests by Mrs. Driscoll's 
Enigma group. 27 The project languished for a 
time, but ideas were formalized in March 1942. 
Machine design began six months later. 28 Hypo's 
task was to help Driscoll's small team make a tra- 
ditional attack on the German Enigma. It was the 
first machine designed especially for the 
American work against the "E" machine. 

(T3//SI) In early 1942, the United States had 
hopes of cracking the Enigma in the same way it 
had broken earlier Japanese cipher machines — 
through methods that included what some called 
"statistical" analysis. Once the entire wheel wiring 
of "E" became known, it was hoped that large files 
(catalogs) could be constructed showing how 
each combination of enciphering wheels would 
"develop" high frequency letters, digraphs, or 
very common words. This "catalog" approach was 
not considered a "cribbing" attack; it was seen at 
the time as a statistical method although counting 
was not required. 29 

fT0//S9 - Constructing catalogs was very labo- 
rious. There had to be a card for each wheel com- 
bination and order, and for each position of the 
wheels. Such catalogs ran to hundreds of thou- 
sands of cards. Some filled an entire wall with file 
drawers. When put into book form, the heavy vol- 
umes demanded along set of shelves. 

ff$//SI) Searching through all the entries to 
find those indicating which wheel settings might 
have produced the enciphered text was also very 
labor intensive. 30 That was why cryptanalysts 
around the world turned to the use of overlay 
sheets. They allowed a speedier and less demand- 
ing way of identifying the possible settings of the 
enciphering machines. 31 But they were limited 
and everyone wanted a faster method. 

(T9//SI) Investing in the construction of cat- 
alogs seemed very wise in 1942 because the 
United States did not have Bombes or the capa- 
bility to continuously find the long and trustwor- 
thy cribs that made the British Bombes so power- 
ful. The Americans did not even have the com- 
mand of the techniques that had allowed Alan 
Turing to apply his indicator- (not crib-) based 
Banburismus IC-like system to "E" since the late 



(TS//SI) The Americans knew that Hypo 
would not be a cure-all machine, but they had lit- 
tle else to rely upon. The enormous amount of 
labor required to prepare Hypo's "database," the 
catalog, seemed worthwhile. Turning the catalog 
into a form that could be used by a high-speed 
machine meant creating a separate roll of film for 
each wheel combination. One was needed for 
each combination of the Enigma's slow and medi- 
um wheels and its reflector. Each frame on a film 
recorded the output of high frequency letters for a 
wheel position. A minimum set of the master 
films was fifty-six, each with over 17,000 
frames. 33 

£ KS//St) The preparation of the message film 
involved as much cryptanalytic persistence as did 
creating the "catalog" films. In one of the most 
common uses of Hypo, ciphertext was partially 
deciphered by pulling out the influence of the pre- 
sumed stecker and fast wheel. 34 Then, the new 
text was put through a crude Letterwriter tape 
machine that recorded the text as tiny dots on 
microfilm. After that, an analyst had to wait the 
many minutes while the film was developed, 


Page 69 

I UH bbCKfcll/lUM I N I JJKbL I U UbA, AUt>, LAW, bBK WW NZU/Al 

dried, and checked for possible defects. 
Producing the message film grew so burdensome 
that IBM and Eastman were ordered to cooperate 
to build a very expensive but labor saving card-to- 
film camera system for Hypo. 35 

(U) The Americans were so desperate for their 
own solution to the Enigma problem in 1942 that 
they did not want to admit to the limitations of 
the Hypo method. Hypo was not powerful. A 
Hypo run needed prior knowledge of the stecker, 
reflector, and wheel order used for an "E" mes- 
sage. With that information it could point out a 
"likely" starting position (window setting) for the 
Enigma wheels. 36 That was all it could do, and 
that was why only two Hypos were built by 
Eastman during the war. 37 Although OP-20-G's 
leaders might have envisioned rooms full of 
Hypos, each running a catalog film against a mes- 
sage, they soon came to treat the Eastman 
machine as only an adjunct to the Bombes. 38 

(U) Hypo looked and behaved much like Icky. 
It used two 35mm films. It was based on dot cod- 

ing and the light bank data entry system. 39 The 
Hypo camera (for dot registration) was an 
improvement over Icky's, however. Data cards or 
tapes signaled which one of the lights in each col- 
umn would be lit, and the tiny dots were record- 
ed on the films with great precision. 

(U) The men working on Hypo also con- 
quered some of the problems of the film stepping 
mechanisms. That allowed a more precise and 
speedy comparison of the films. When the catalog 
and message films were placed on the machine, 
one was held stationary while the other flashed by 
it. As in Tessie and Icky, when the second film 
completed a revolution, the first was stepped one 
increment. That took less than five seconds. 

(U) The "statistical" test in Hypo was a 
desired level of coincidence between text and 
master film spots. As in Tessie, Hypo's photocells 
monitored a zone rather than an individual col- 
umn. The likely enciphering-wheel positions 
were identified simply by enough light reaching a 
photocell. When the machine stopped, its opera- 
tor wrote down the location of the hit. 

(U) Hypo was an analog 
machine designed to locate. It was 
not coaxed to tally until the end of 
the war. Even then, it remained a 
very simple device. Despite that, 
Hypo proved as useful as Tessie 
did, though neither machine solved 
any systems by itself. Copies of 
Hypo were supplied to the army's 
cryptanalysts, and a second and 
more complex version was con- 
structed for OP-20-G later in the 
war. By early 1945, Hypo was also 
being used against Japanese sys- 
tems, after it had undergone some 
significant modifications. 4 ° 

Page 70 

woMsusmsmmmm miiri . mwrSftH pnni " n>l '" l " f4 


(U) A Paper War, Perlmps 

(U) Hypo did not seem a winner in 1943, how- 
ever. The delays in delivering Icky and Hypo had 
made Meader and Wenger fearful that Eastman 
would be unable to produce any device except the 
crude analog and plate IC Machine. Tn the critical 
first two years of the war, they also feared that 
IBM would not deliver its promised data conver- 
sion machines. In addition, there were signs that 
the next model of the Bush Comparator was in 
serious trouble. At the beginning of the war, OP- 
20-G was hedging all its technological bets. 
Although Howard had advised against a paper 
tape machine, the navy ordered him to stop his 
exploration of microfilm and draw up the essen- 
tials of an upgraded paper tape Bush Comparator. 
Howard helped draft a sketch of a slightly revised 
version of the old Comparator and sent it to the 
bureau's contractors, NCR and Gray Electric. 41 
NCR and Gray set out with a great deal of enthu- 
siasm, and the navy looked forward to a third ver- 
sion of the Comparator in a few months. 

(U) Using the older 70-millimeter paper tape, 
but with room for thirty-two rather than twenty- 
six characters, the new Comparator tallied and it 
employed parallelism. It was able to handle and 
record up to five pattern tests at once. Its circuits 
and plugboards were more complex than the ear- 
lier model, and it was given an important new 
capability: it could locate. One tape could be held 
stationary while the other sped past it. The sta- 
tionary tape then moved one increment for 
another pass of the second tape, stopping when a 
match was sensed. 

(U) Four copies of the new paper tape 
Comparators were constructed between 1943 and 
1945. 42 They seemed so promising when they 
were first designed that Britain asked for two. 
Later, its codebreakers decided against the 
machines, and the two were sent to the army's 
men at Arlington Hall. 43 

(U) The World War II paper tape. Comparator 
proved an essential tool for the jobs that needed 
tallying, but, unfortunately, the new machines 
could not be convinced to run appreciably faster 
than the 1938 Comparator. The 1943-1945 mod- 
els continued to have a relatively slow speed, 
eighty-five characters a second. 

(U) One irksome feature of the Comparators 
was corrected by the end of the war. Like the orig- 
inal Bush Comparator, the 1943 device printed 
every result, appreciably slowing its performance. 
To speed it, an electronic circuit was added that 
allowed printing only when a highly improbable 
level of coincidence was computed. 44 

(U) The task of making a reliable punch was 
probably turned over to a Bass River, 
Massachusetts, firm. But even an expert private 
manufacturer could not overcome the punch's 
difficulties. The Oano Company had a tough time 
with the design and soon separated itself from the 
Comparator project. 45 

(U) By late 1942, the Comparator seemed des- 
tined to fail again. There was too much for 
Meader and Howard to keep under control. 

(U) The Comparator Dies, Again 

(U) John Howard spent the first months of 
1942 traveling from place to place with Ralph 
Meader trying to force progress on the Eastman 
and NCR-Gray machines. By mid-1942 Meader 
and Wenger sensed that something was wrong 
with the Gray-NCR-MIT effort. When all the 
components were finally delivered to 
Washington, they did not fit together. The situa- 
tion was so bad that the Comparator was 
returned to New York where the Gray and navy 
engineers redid most of its parts. 

(U) When the Comparator was finally sent 
back to Washington, the navy engineers had to 
spend a great deal more of their valuable time 


Page 71 


reworking the comparing and counter-printer 
units. That further delayed putting it to use. 46 

(U) The new Comparator did not go into 
operation until November 1943. 47 

(U) By then, the relations between Gray and 
OP-20-G had become quite tense. What the navy 
interpreted as disorganization in New York had 
much to do with its alienation. 

(U) One explanation of why only four copies 
of the new Comparator were built is the difficulty 
"G" had controlling Gray's work. The lack of con- 
trol became quite evident in mid-1944 when Gray 
Manufacturing took out a full-page advertise- 
ment in a widely read electronics journal. It 
showed the world what kind of tasks Gray had 
been doing for the government that had earned it 
an Army-Navy E award. The bottom third of the 
advertisement caused an emotional outburst in 
Washington. The last two items on the page 
stated: 48 

(I I) < Optical work including the design and con- 
struction of various units in the projection field 
including photographic technique, motion pic- 
ture, and optical systems involving condensers, 
prisms, and associated reflector equipment... 
communication equipment, electrical counting 
and calculating devices, including communica- 
tions devices for producing or operating from 
perforated, inked, and ended tapes of various 

(U) The advertisement infuriated the crew at 
OP-20-G. On top of all the manufacturing prob- 
lems, Gray had endangered the security of "G's" 
RAM program. Larry Steinhardt could not con- 
tain himself when he read the advertisement. He 
tore it out of the journal and immediately sent it 
to Howard Engstrom with a message he wrote on 
it that said, "Note below an excellent description 
of the 70mm junkpile this outfit built. Please pass 
to Meader." Although the way the contractors had 
organized their work had much to do with the 

Comparator's problems, the underlying cause of 
all of the difficulties was stubborn technologies. 
They made it impossible for the nation's best 
engineers to fulfill Bush's promises. 

(U) The machines built between late 1943 and 
the end of the war had to be retreats from Bush's 
visions. The Copperheads, for example, had to be 
compromises between an engineer's pride and 
cryptanalytic needs. Other machines, such as 
Bulldozer and Duenna, were advances on the 
state of the electronic art, but they were based on 
ideas and techniques that were very different 
from those Bush had championed. 

(V) Almost Another Digital Machine 

(U) The other major attempt by the navy's 
team to fulfill Bush's promises was the 
Copperhead series. 49 Several different 
Copperheads were designed, and five copies of 
one of the series were built under Lawrence 
Steinhardt's direction at National Cash Register 
and at "M's" Washington engineering laboratory. 
Constatction began in late 1943. Unfortunately, 
all of the more ambitious plans for the 
Copperheads had to be put aside because of tech- 
nical problems and cryptologic emergencies. 
Only the copies of the rather simple Copperhead 
I were built. 

(U) In 1943 the Atiantic crisis eased some- 
what, giving "M" a bit of time to turn to Japanese 
problems. Lawrence Steinhardt was assigned the 
job of designing Rapid Machines to attack addi- 
tive systems. Additive systems were codes with 
random numbers added or subtracted from the 
underlying numeric codes. Among many others, 
the major Japanese naval codes used additives. 
The fleet operational code, JN25, was of very spe- 
cial importance to American intelligence. But it 
had proven to be a very difficult adversary, espe- 
cially because the Japanese frequentiy changed 
the long list of additives used to superencipher its 

Page 72 



(U) Discovering those addi- 
tives was a tedious process. "G" 
had to call on many different 
methods of attack. IBM equip- 
ment had been modified to speed 
the identification of the superen- 
cipherments, but the process 
remained very slow and seemed in 
need of Rapid Machines. In 1943 
it was decided to start an additive 
RAM program. Following the new 
habit of using the names of snakes 
for Japanese problems, the proj- 
ect was called Copperhead. 

(U) Still excited about optical- 
electronics, Steinhardt prepared 
the outlines for at least five differ- 
ent devices for the Copperhead 
problems. In his plans, the more 
complex models were to be able to 
add and subtract and to test statis- 
tical weights at electronic speeds. 

(T0//CI) Copperhead II, for 
example, was designed to be able 
to add clusters of additives to message text, then 
compare the results against along list of known 
high-frequency code groups. 

^-Copperhead V was a truly grand vision. If 
it had been built, it would have been twenty feet 
square. It would have had to have been that big to 
be able to match strings of additives against 
cipher text, then perform a true statistical test for 
nonrandom letter frequencies. 50 That called for 
sophisticated electronics and very high-speed 
input. The complex job assigned to "V" seemed to 
call for microfilm for input and perhaps for a vast 
memory. But Steinhardt was aware of the prob- 
lems at Eastman, and at the onset of the 
Copperhead project he decided that the older 
punch tape approach would be best. 

(U) More than a year was spent searching for 
a new tape and designing a revolutionary punch. 

After testing many materials, including alu- 
minum foil, a 70mm opaque polystyrene tape was 
selected. It had the stability needed for very high- 
speed transport past the scanning station and did 
not distort when there were humidity changes. Of 
great importance, it could accommodate a data 
density about twice that of the Comparator's 
paper tapes. 

(U) The Copperhead punch was a major engi- 
neering feat. Its main cabinet was over six feet tall 
and was wider than a phone booth. It was packed 
with delicate mechanical and electronic parts that 
perfectly aligned two tapes and then punched a 
reciprocal code. Each column on the tape had 
room for twenty-five tiny dots for message char- 
acters and several others for identification of the 
message. The punch was designed around the 
blackout system. Learning from earlier microfilm 
explorations that the absence of light was easier 


Page 73 


to monitor than its presence, one tape was 
punched to be the complement of the two-of-five 
code on the other. The designers were so pleased 
with the Copperhead's punch, they built modified 
versions of it for the older Gray-NCR 

(U) Copperhead I used two sets of sophisticat- 
ed motor-driven reels. It had a sensor system to 
manage the end-of-tape condition and the 
mechanical components needed to automatically 
rewind and step the tapes. Also, the machine was 
a landmark in optical sensing. It was built to scan 
one hundred message columns at a time! 

(TJ) As with Icky, the Copperhead team had to 
take some significant backward steps to produce 
a machine to meet the war crises. Only one ver- 
sion of the Copperheads was built, and it was 
unable to count; it simply located message 
groups. As many as five of the Copperhead I 
machines were constructed and in operation by 
the opening of 1945. But they were very limited 
punched tape versions of the IC Machine and 

(V) The Old Technologies Are the Best 
Technologies, for a Time 

(U) In the spring of 1942, the Copperheads 
were not yet well-formed ideas, and all other 
Rapid Machines were in trouble. Even the refur- 
bished 1938 Comparator, the only working Rapid 
machine, was not proving its worth. Lawrence 
Steinhardt had to strip it of many of its original 
functions to make it reliable enough for use in 

4S* By October, Steinhardt had built a crude 
prototype and had drawn the outlines for a much 
more sophisticated machine to identify Japanese 
code groups based on frequency criteria. 

■48) For the emergency machine, 700 of the 
most frequent groups were stored on film in 
descending order of frequency. The meaning of 

the group and its known relative frequency were 
listed next to the group's number and language 
equivalent. The "selector" was a simple relay store 
with "pin" settings indicating the frequency of the 
various code groups in the message being ana- 
lyzed. When the message group and its frequency 
matched the composition of a group on the mem- 
ory film, the film's entry was recorded by a fast- 
flash system. After the run, the new film was 
developed and sent to an analyst who used the 
information to help decrypt the message. 31 

(TS//SI) After the first lash-up came a series 
of increasingly complex "Full Selectors." By the 
end of 1942, the first model had been modified 
through the addition of more sophisticated relay 
boxes; and by that time there were plans for a 
much larger and powerful device, Mercury. 

(TS//8I) Although the hopes for a huge elec- 
tronic version were defeated, Mercury became a 
room full of relay racks that performed a sophis- 
ticated "weighted dictionary look-up" test to iden- 
tify code groups. 52 Unfortunately for the navy, 
Mercury was not working until the summer of 

(U) Meanwhile, the Tabulator's Revenge 

(U) While Wenger worried about the absence 
offunctioning Rapid Machines, those who had 
advocated the development of older technologies 
seemed to be vindicated. The old timers were in 
charge of tabulator development, and in 1942 
they were the ones delivering cryptanalytic 
results. 53 

(U) IBM sent all the tabulators and sorters 
and collators OP-20-G could make room for, and 
the company began to create a host of very pow- 
erful additions for its machines. After "G" moved 
to its new quarters at an elegant girls' school on 
Nebraska Avenue and had adequate space, OP- 
20-G became one ofthe world's largest users of 
TBM equipment. "G's" TBM machines were count- 

Page 74 

jpp cEPBtTiirflyiiiTTinri Tn i ni m i r ,uu i .mj inumu;i))«l 


edin the hundreds, and they used millions of 
punch cards a week. 54 

(U) Acquiring standard IBM machines was 
relatively easy. Alone among almost all business 
machine manufacturers, IBM had been permitted 
to continue manufacturing its products during 
the war. Its "tabs" remained stock items, and OP- 
20-G already had high priority status. 

(U) But gaining IBM's commitment to contin- 
ue to alter its machines (or to allow OP-20-G to 
do so) proved more difficult. Joseph Wenger had 
to make a personal visit to Tom Watson to con- 
vince him to grant OP-20-G's requests special 
attention. By the end of the war, IBM and the 
armed services' engineers, many of whom were 
drafted from IBM and were sent directly to 
Washington, had created modifications that 
allowed the electromechanical machines to per- 
form all the cryptanalytic functions. Because of 
those modifications, IBM's equipment remained 
the foundation ofOP-20-G's operations through- 
out the war. 

20-G/Yard crew did not demand the creation of 
an all-purpose tabulator or a general-purpose 
relay computer, but they asked for some chal- 
lenging engineering advances. The requests indi- 
cate the old-timers had long had their own alter- 
natives to Bush's mid-i930s Rapid Machine pro- 

(U) As well as the special electromechanical 
attachments for OP-20-G's tabulators, IBM creat- 
ed ambitious relay additions. The new IBM 
devices were better able to identify and tally par- 
ticular code groups and to search for repetitions 
of character patterns. Among the more ambitious 
proposals for IBM equipment were the Navy 
Change (NC) machines. 

(U) The NC machines were more than stan- 
dard tabs with a few additions hung on them. 
Some of the thirteen types of Navy Change 
machines came close to being special relay com- 
puters. Others had special high-speed electro- 
mechanical accumulators and some had electron- 
ic tubes. 57 

(U) With the outbreak of war, the tabulator 
group at "G" was able to expand and to convince 
IBM to produce specialized equipment. IBM and 
the navy began a cooperative effort that lasted 
throughout the war. 55 A number of IBM men 
went to Washington, and a host of new attach- 
ments were developed. Some allowed more effi- 
cient additive stripping. New devices provided 
more effective multiple key sorting and the offset 
and comparison of messages for IC analysis. The 
location of code words was made faster by other 
additions to the tabulators, sorters, and punches. 

(U) Although IBM played an important role in 
OP-20-G's war, it was not asked to take a signifi- 
cant part in the Rapid Machine program. 56 One 
reason for not calling on IBM was that OP-20-G 
was already asking a great deal of the company. 
In 1942 the requests by the tab group at"G" for 
electromechanical and relay devices were enough 
to keep the company's best men busy. The OP- 

(U) IBM's Most Special Contribution 

(U) During 1941 and early 1942, before 
Engstrom's group gained real power over 
machine development, and as the Eastman and 
Gray-NCR projects were faltering, IBM and the 
men at"G" created another innovative system, 
the Letterwriters. Those devices brought OP-20- 
G's data handling into the modern era because 
they linked teletype, tape, card, and film media. 
The Letterwriter system tied special electric type- 
writers to automatic tape and card punches and 
eventually to film processing machines. 

(U) Before the war the radio intercept person- 
nel wrote out the messages they heard on forms, 
then forwarded them by mail or keyed them as 
telegrams. Because OP-20-G had just begun to 
develop teletype and radio networks, it took 


Page 75 


weeks to send all but the most vital messages 
from the Pacific. 58 

(U) There are somewhat conflicting stories 
about the origins of the Letterwriter (CXCO) 
equipment, perhaps because its prehistory was 
linked to so many different groups within the 
navy. The timing is not entirely clear, but some- 
time in late 1940 Hooper's previous connections 
to the man who had sold his advanced electric 
typewriter business to IBM led to some interest in 
perhaps modifying his machines to turn them 
into data processing devices. 

(U) The interest was turned into action in 
early 1942 when a young IBM engineer entered 
the navy and was assigned toOP-20-G. John 
Skinner had worked on a special typewriter-tele- 
type project at IBM. When he had enough experi- 
ence to appreciate "G's" data processing prob- 
lems, he contacted his ex-boss and arranged to 
have some equipment shipped to Washington. 59 
After IBM engineers arrived with the devices and 
demonstrated their potential, there was an imme- 
diate request that IBM launch a major project. 

Within less than a year, the first production 
Letterwriter devices were delivered to the crypt- 

(U) The timely appearance of the first 
machines was a result of IBM's earlier commer- 
cial efforts at its Electromatic division. 60 The sys- 
tem centered on a special electric typewriter, a 
tape punch, and a tape reader. It was hoped they 
would eventually allow the creation of machine- 
ready data directly from "G's" new international 
telegraph system. 

(U) The Letterwriters were not intended to be 
analysis machines, but to fill the gap left by the 
delayed RAM program. The engineers in 
Washington turned the Writers into much more 
than data entry devices. By adding simple plug- 
boards, the engineers made the machines pro- 
duce worksheets for the cryptanalysts and change 
one code into another. 61 By 1942 the Letterwriters 
were evolving into machines for analysis. First, 
the typewriters were modified to allow the print- 
ing of more sophisticated worksheets. Flugboxes 
were added which allowed complex substitutions 

Page 76 



of one character for another. This helped deter- 
mine the settings of the letter-changing plug- 
boards on encryption machines. In addition to 
being useful for the analysis of steckering, the 
modified Letterwriters helped to strip cipher 
wheel patterns from messages. 

(U) Simple changes made the Letterwriter 
equipment useful for another very important but 
time-consuming task, the analysis of wheel set- 
tings. When an analyst thought he had found the 
correct combinations on an enemy system, he 
would set up a copy of the encryption machine's 
wheels, lugs, and plugboards and type in parts of 
the encrypted message. He then examined the 
output to see if it was sensible. By coupling a 
Letterwriter tape-reader to one of the American 
copies of a foreign cipher machine, an analyst 
would not have to repeatedly enter a message 
through the machine's keyboard. 

(U) In the Absence of Rapid Machines 

(U) The delays in the delivery of the Rapid 
Machines led to another use for the Letterwriters. 
The Yard's men decided to build more far-reach- 
ing extensions of them. The first of their 1942 cre- 
ations was a frequency counter. Aptly titled The 
Simple Frequency Counter, it was among the first 
of the new machines to 
be delivered to OP-20- 
G. The Simple Counter 
and its descendants 
had a power Bush's 
machines did not pos- 
sess: they were able to 
recognize and record 
individual letters. The 
recognition, counting, 
and recording of partic- 
ular letters and poly- 
graphs demanded too 
many complex elec- 

tronic circuits and parts for computer technology 
of the early 1940s. 

(U) The Counter saved preparing IBM card 
decks and the many steps involved in repeated 
sorting. It was such an effective design that in 
1943 a grand extension of the Counter was con- 
structed at NCR. The NCR machine, Mike, tallied 
digraphs. Despite the low speed of such devices as 
Mike, the inability to deliver any Rapid Machines 
led the Yard's men to create yet another type of 
relay-electromechanical analyzer. They designed 
a machine, Mathew, 62 to perform additive strip- 
ping. Like the Counter, the Mathews proved reli- 
able and were used throughout the war. Mathew 
was so rugged that it was applied to more than 
traditional stripping. 63 It was used on such jobs 
as removing the influence of a cipher wheel from 
an encrypted message. Over the years, the many 
Mathews (at least four were constructed) proved 
useful against a majority of the encryption sys- 
tems attacked byOP-20-G. Mathew was not a 
general-purpose machine, however, and its tech- 
nology dated from the 1920s. Its processing 
power was limited by the speed of its tape readers 
and its typewriter. But it was able to perform 
faster than the tabulators and to fulfill functions 
too complex for the electronic Rapid Machines of 
the era. 



ieewcweoM i N r /fren. touoa, ami, pah, opn u na i eu/a i 

CL9 iVotes 

l.-fS). Almost all the documentation on the first 
year of the Eastman work and its first machines has 
been lost. For an insight into the problems of rushing 
into development and lack of coordination among the 
Eastman teams and the navy, see Rowley's comments 
about his mid-1943 tour of the Eastman projects, (S) 
NSA CCH Series XII Z, "Inspection of RAM Under 
Construction at NY and Rochester," 31 May 1943. 

2.(U) The estimate of when the IC machine was 
ready is based on very circumstantial evidence. But it 
is clear that it was in use well before any other Rapid 
device, including the American Bombe. NSA RAM 
File, Report ofR. LMeader, Captain USNR, to J. H. 
Wenger, Captain, USN, "14 Days Training Duty, 
Report of," January 21, 1949, and Communications 
Intelligence Technical Paper La, "Technical Report: 
The Index of Coincidence Machine" March 1945. 

3- (U) Typically, there were other precursors of the 
IC machine, including patented devices intended for 
business applications. See, for example, H. Soper, U. S. 
Patent 1,351,692, August 31, 1920. 

4-(U) NSA RAM File, 0P-20-G toOP-20-A 
"Meeting with Prof. Howard," November 5, 1941, and 
Communications Intelligence Technical Paper l-a, 
"Technical Report; The Index of Coincidence 
Machine* March 1945. (T5?fS4)Jhe first of the plate 
IC machines was delivered in August 1942. But it 
needed some fine-tuning and then had to be used in a 
controlled area rather than, as planned, at the crypt- 
analysts' desks. The device was redone several times 
before the end of the war, and the army group at SIS 
used several copies. Eventually, it adapted to the use of 
film as well as plates. pe^NSA CCH Series XII Z, 
Herbert W.Worden, "EDP Machine History." 
XTSfmi- CCH Series XII Z, LeRoy H. Wheatley, 
"Cryptanalytic Machines in NSA," 30 May 1953, and 
various years. 7TS7 l 7'SI)JXH Series XII Z, copies of 
various MAC Outlines, circa 1953. (TS//SI) CCH 
Series XII Z,(S 12008) Navy Dept., Office of Chief of 
Naval Operations, DNC (OP-20-G), RIP 425, "The 
American Attack on the German Naval Ciphers," 
October 1944 [sic]. 

5^&L. AHA ACC 1890, February 27, 1943, 
"Accuracy of the I.C. Machine." AHA ACC 1890, "New 

I.C. Equipment and Alterations Made on Old 
Equipment." AHA ACC 1890, August 2,1943, 
"Electronic Use of I. C. Projectors." AHA ACC 1890, D~ 
GM-5 toGM-5 "Changes Made on I.C. Reader and 

6.(Tfl//0I) On the confusion over the first of 
Eastman's film machines, Tessie, S409, Brief 
Descriptions of RAM Equipment, Navy Dept, 
Washington, D. C, 1947, and Leroy Wheatley, Brief 
Descriptions of Analytic Machines, NSA 34, 1954. 
AHA ACC 1890, GM-2 to G-50, "Tetra Projector #2 
(RAM-5): Name for." AHA ACC 1890, February 27, 
1943, "Accuracy of the I. C. Machine." 

7-^NSA CCH Series XII Z, NSA OH 16-85, Oral 
History Interview with Capt. John A. Skinner, 25 
September 1985, 24. 

8.m NSA CCH Series XII Z,J. A Skinner, 
"Proposal for Decoding Device," OP-20-GM, 16 
February 1943. 

9. (U) Four-character code groups were used in 
important German and Japanese systems. It is not 
known ifTessie was originally built for use against 
both of them. The Japanese high-level fleet code used 
a four-digit code. The very important U-boat short sig- 
nal code was used to flash location messages and was 
tapped by the Allies for cribs. The short signals were 
also used as cribs into the four-wheel Enigma systems. 
Tessie was modified later in the war specifically for the 
German short signals. RAM File, History of OP-20-G 
/NCML/4e, 106. 

lO.-fS^AHA ACC 1890, OP-20-GM-10 toOP-20- 
GN, January 23, 1943, "Ram-2, Improvements on 
Performance Of." AHA ACC 1890, M-4 toCM-5 
March 6, 1943, RAM-2, "Changes in Operation of." 
AHA, ACC 24880, CIT Technical Paper 9, Tessie SS, 
Vol I, CNO, Navy Dept, Washington, D.C., May 1945. 
AHA ACC 1890, "RAM-2 Operating Procedures." 

11. (U) Near the end of the war, counting circuits 
were added to the device, making it a weak version of 
a microfilm Bush Comparator. NSA RAM File, History 

12. e»*-AHA, ACC 24880, CIT Technical Paper 9, 
"Tessie SS, Vol I," CNO, Navy Dept., Washington, D.C., 
May 1945. 

Page 78 


Tor ? r rnETfr?r ,IIIT " nri Tn " rl " In PA " ' >rmflupmgu>x i 

13. ■$$). AHA, ACC 1890, Special Applications 
Section, Bureau of Ships, to OP-20-GM, August 11, 
1943, "Equipment Developed byEK Co." 

14. 4$). AHA Ace 1890, "Report on Enigma Test 
Run on RAM-2, January 7-8, 1943-" AHA ACC 1890, 
OP-20-GM-10 toOP-20-GM, January 23, 1943, 
"RAM-2, Improvements on Performance of." AHA 
ACC 1890, GM-4 toGM, June 23, 1943, "RAM-2, 
Technical Details of recent work on." AHA ACC 1890, 
GM-4 to GM, June 30, 1943, "Ram-2, Comments on 
Performance of and July 20, 1943, "RAM-2 Camera 
#4, comments on design of." 

15.-(¥S)-NSA AHA ACC 1890, GM-4 toGM, June 
23, 1943, "RAM-2 Technical Details of recent work 

16. ■&) NSA AHA ACC 1890, GM-10 toGM, 
January- 23, 1943, "RAM-2, Improvements on 
Performance of." AHA ACC 1890, GM-4 to GM, "24- 
hour trial run in E traffic using RAM-2." 

17. (U) Letters from Joseph Eachus circa 1988. 
Near the very end of the war, counting circuits were 
added to the device, making it a weak version of a 
microfilm Bush Comparator. But until then it did not 
even record the place where a "hit" occurred. NSA 
RAM File, History- of 0P-20-G/NCML/4e. 

18. (S//S*} NSA RAM File, Report of R. I. Meader, 
Captain USNR to J. N.Wenger, Captain, USN, "14 
Days Training Duty, Report of," January 21, 1949. On 
Tessie's rebirth as the Symmetric Sequence Machine 
in 1944, <& NSA CCH Series XII Z, RAM list and 
Conference at Dayton, 11 April 1945, (XS}- NSA AHA 
ACC 24880, C1T Technical Paper 9, "Tessie SS, Vol I," 
CNO Navy Dept. Washington, D.C., May 1945. AHA 
ACC 1890, GM-2 to G-50, May 25, 1944, "Tessie: More 
Complete Conversion to symmetrical sequence work." 

19. (TS//S } ) On the meaning of the terms, NSA 
CCH Collection, "Army-Navy, Descriptive Dictionary 
of Cryptologic Terms," Headquarters, Army Security- 
Agency, February 1947. 

20.-PS} NSA CCH Series XII Z, "Brief Description 
of RAM Equipment," Navy Dept. Washington, D. C, 
October 1947, 37. 

21. -(S}- NSA CCH Series XII Z,A. W.Tyler, 
"Tetragraph Machine II," (ICKY) 21 February 1944. 

22. (T3//3I) NSA CCH Series XII Z, "Office of 
Computers, List of Computers," nd.'^NSA CCH 

Series XII Z, "ICKY," circa 1944- TTS7W NSA CCH 
Series XII Z, "Hypo I - Hypo 111," March 1954- 

23. (U) NSA RAM File, "MAC. Outlines #17, 
70mm Comparator," April 1947- The German inven- 
tor, Goldberg, had chosen the blackout methods. 
Emanuel Goldberg, U. S. Patent 1,838,389, Statistical 
Machine, December 29, 1931, Filed April 5, 1928. 

24. (U) NSA RAM File, Communications 
Intelligence Paper 6, ICKY, Washington, D.C. April, 


25.4 TS//0 fl Hypo was delivered toOP-20-G in 
October 1943, just as the bombes became operational. 
( TS//SI) NSA CCH Series XII Z,"Hypo I- Hypo III," 
March 1954. 

26. (TS//SI) NSA CCH Series XII Z, "Hypo I - 
Hypo III," March 1954. 

27. (U) NSA RAM File, "List of Equipment for 
Enigma Problems." Note that high-level policy had led 
the navy to place little emphasis on Hypo during 1941. 
Howard was told that Mrs. Driscoll's problem was "not 
that important" and to place emphasis on other 
machines. NSA RAM File, November 14, 1941, Bureau 
of Ships to Howard, "Driscoll's problem not that 

28. (TC//SI) NSA CCH Series XII Z, (S12008) 
Navy Dept., Office of Chief of Naval Operations, DNC 
(OP-20-G), RIP 425, "The American Attack on the 
German Naval Ciphers," October 1944. [sic], 51. 

29. (TS//6I) Hypo's initial outlines contained an 
explanation of how it might be constructed so as to be 
used as a true crib device. (TS//SI) NSA CCH Series 
XII Z, CNO CIT Technical Paper TS-10/E-3, "Enigma 
Series: Vol. #, Statistical Studies," January 1946. 

30. (TB//SI) NSA CCH Series XII Z,(Si20o8) 
Navy Dept., Office of Chief of Naval Operations, DNC 
(OP-20-G), RIP 425, "The American Attack on the 
German Naval Ciphers," October 1944. [sic], 117. 

31. (U) Britain also had statistical methods, such 
as Banburismus, which brought forth some ideas 
about a film machine in England, perhaps as early as 
1939. England may have built film devices that 
equaled or exceeded those built in the United States 
during the war. Andrew Hodges, Alan Turing: The 
Enigma (New York: Simon and Schuster, 1983), 178, 


Page 79 


32-4¥S) NSA CCH Series IV.7.20, A. P. Mahon, 
The History of Hut Eight, 1939-1945-" On the state of 
American knowledge of Enigma methods at the out- 
break of the war, (T0//QI) NSA CCH Series XII Z, 
(S12008) Navy Dept., Office of Chief of Naval 
Operations, DNC (OP-20-G), RIP 425, "The American 
Attack on the German Naval Ciphers," October 1944. 

33. (TS//SI) NSA CCH Series Xli Z, CNO CIT 
Technical Paper TS-10/E-3, "Enigma Series; 
Statistical Studies," January 1946, E3-12. 

34. (TS//SI) Two different uses of Hypo are 
described in the existing literature. For the one 
described here, see (TS//SI) NSA CCH Series XII Z, 
"Hypo I - Hypo III," March 1954, and for the other 
more crib-like description see, TTStVSJI NSA CCH 
Series XII Z, CNO CIT Technical Paper TS-10/E-3, 
"Enigma Series: Statistical Studies," January 1946. 

35. (TS//SI) NSA CCH Series XII Z, "Hypo I - 
Hypo III," March 1954. 

36. (•€> NSA CCH Series XII Z, Descriptions of 
NSA Early SPDs and Computers, as compiled from 
various NSA sources. (TS//SI) NSA CCH Series XII Z, 
LeRoy H. Wheatley, "Cryptanalytic Machines in NSA," 
30 May 1953. and various years. (TS7VW) NSA CCH 
Series XII Z, copies of various MAC Outlines, circa 


37. (ۥ) However, Lawrence Steinhardt completed 
another in 1946 and a fourth in the early 1950s. (C) 
NSA CCH Series XII Z, Descriptions of NSA Early 
SPDs and Computers, as compiled from various NSA 

38. © Hypo was modified for use against the 
Japanese 157 Jade machine.'TjS^NSA CCH Series XII 
Z,H. H.Campaigne "Use of Hypo on the JN-157," 21 
February 1944. 

39. (U) NSA RAM File, CNO, USNC, CITP TO-33 
"Overhaul of Hypo #1," Washington, D.C., June 1945- 
Letters to author from Joseph Eachus. Microfilm and 

40. (U) NSA RAM File, W. A. Wright to OP-20-G 
February 21, 1944, "Comparison of Army and Navy- 
Enigma Equipment." NARA RG457, SRH-200, 
"Army-Navy Collaboration 1931-1945/ 216-8. For 
later models and use against Japanese systems: NSA 
RAM File, June 16, 1947, OP-20-G Research 

Committee Meeting; January 5, 1945, "Hypo Stepping 
Switch"; "History of OP-20-G /NCML/4e"; and CNO, 
U.S. Naval Communications, CITP TO-24 "JN-37 
Problem on Hypo," Washington, D.C., May 1945. 

41. (U) Hagley Museum and library, Accession 
1825, Honeywell vSperry-Rand, Trial Records, 
Deposition of Joseph Desch. NARA Suitland, OSRD 
Contract Files, OEM-275 November 28,1941, "NCR- 
MIT counters." NSA RAM Files, Joseph Desch to 
OSRD, February 12,1943, "Only Navy work at NCR." 

42. (U) The estimates of the number of 
Comparators built during the war vary from six to as 
many as twenty-eight. Four is the correct figure. The 
reason for the high estimate was probably that all the 
later postwar Comparator-like machines were includ- 

43.4S) Office of Naval Research, Patent File on 
"Electronic Comparator, Vannevar Bush." V. Bush, 
U.S. Patent, February 17, 1959, "Electronic 
Comparator," 2,873,912. Of importance for the post- 
war history of the Rapid Selector, the Comparators 
became the basis for the navy's patent claims over 
optical-electronic devices. {§} On the British and army 
comparators, 65} NSA CCH Series XI E, Hagelin, Box 
2, Folder, "Comparators.." 

44. (8) On the rare event circuit, (S) NSA CCH 
Series XII Z, J. H Howard, "70MM Comparator & Rare 
Event Circuit," 27 October 1944. 

45. (TS//SI) NSA CCH Series XI E Hagelin, 
Box,"Notes on various topics." 

46. (£) NSA CCH Series XII Z, "Inspection of RAM 
Under Construction at NY and Rochester," 31 May 
1943. m NSA CCH Series XI E, Hagelin, Box 2, 
Folder, "Comparators." 

47. (U) NSA RAM File, Report ofR. I. Meader, 
Captain USNR to J. N.Wenger, Captain, USN, "14 
Days Training Duty, Report of," January 21, 1949- NSA 
RAM File, CNO, U. S. Naval Communications, CITP 
TS "Machine Descriptions," Washington, D.C., circa 
1945. "Mike, Comparator." NCML-CSAW Message 
File, April 14, 1944, "Punch being modified at Gray." 
The Americans were not the only ones to have prob- 
lems with tape machines. Britain's attempts to create 
similar machines, the Robinsons, faced even greater 
difficulties. As the new Bush Comparator was going 
into operation, Britain was still testing its first two tape 

Page 80 



systems and would soon turn away from such devices 
because coordinating the tapes was too difficult. Allen 
W.M. Coombs, "The Making of Colossus" Annals of 
the History of Computing 5(1983): 254- Brian 
Randell, "The Colossus," in N. Metropolis et al. (ed.), A 
History of Computing in the Twentieth Century, (New 
York, 1980), 47-92. 

48. m NSA CCH Series XII Z,Gray 
Manufacturing Co., "Design Advertisement," June 


49. (U) NSA NCML-CSAW Message File, mes- 
sages to and from Dayton and Washington, November 
1943 to March 1945. NSA RAM File, "Final Report, 
Copperhead II," Communications Intelligence Paper 
24, and Communications Intelligence Paper 41, 
"Copperhead I Punch and Copperhead I Scanner." 

50. -m NSA CCH Series XII Z, RAM list and 
Conference at Dayton, 11 April 1945. (S) Steinhardt, L. 
H., "Copperhead II (Project M-230) Final Report," 9 
November 1944-TfSiJMSA CCH Series XII Z,"Use of 
RAM on Jap Naval Problems of BII type," 9 June 1944. 

51.4S) NSA CCH Series XII Z,L. R. Steinhardt, 
"Full Selector," 31 October 1942. 

52.4S) NSA CCH Series XII Z.L. R. Steinhardt, 
"Full Selector," 31 October 1942. FWStLNSA CCH 
Series XII Z, "Office of Computers, list of Computers," 

53, ( TS//SI) The group of practical engineers were 
probably the ones who built the rather crude but use- 
ful electromechanical Shinn and Ely machines during 
1941 and early 1942. (S) NSA CCH Series XII Z, OP- 
20-G War Diary, OP-20-GS, Machine Processing, 
February 1942-January 1945. (TS//SI ) NSA CCH 
Local Archive, "Army-Navy Descriptive Dictionary of 
Cryptologic Terms," Army Security Agency, February 
1947. The descriptions of these machines were not 
located. IBM aids OP-20-G, 1942. 

54. (U) NARA RG457, SRH-349, "Achievements of 
the SSA In World War H," 18. In January 1941 OP-20- 
G Washington had 16 IBM machines, in 1945, some 
200. NARA RG457, SRH-197, "US Navy 
Communications Intelligence, Organization, Liaison 
and Collaboration 1941-1945." University of 
Pennsylvania, Van Pelt Library Archives, Papers of 
John Mauchly, October 11, 1944, "Mauchly notes on 
meeting with Kullback of ASA" 

55. (U) NSA RAM File, OP-20-G to Radio Sound 
Branch, September 5,1941, January 16, 1942. NSA, 
Tabulating Machine File, July 24, 1941 and December 
6,1941 to Radio Sound Branch, Design Division, 
Bureau of Ships. On frictions with IBM, CNO to BuEng 
1-3-24; "Conference With IBM," May 23, 1934- 

56. (U) NSA, Tabulating Machine File, OP-20-G to 
BuShips, July 24, 1941. 

57.f¥9tNSA CCH Series XI E, Hagelin, Box 2, "NC 
Machines." The various types of NC machines were 

NC 1: consecutive numbering device 

C2: relay adder to mechanize decryption of addi- 
tive cipher 

NC3: "single eliminator," which selected duplicate 
cards in a deck leaving only unique ones; it used vacu- 
um tube circuits and read 300 cards a minute (NC 12 
replaced it). 

NC4: selective punch whose relay additions 
allowed a variety of substitutions to be punched on 


NC5: pattern punch whose abilities included 
searching for isomorphs 

NC6: column differencer whose amazing accumu- 
lator could hold up to 400 items, recognize the largest, 
and punch an indicating card. It also matched high fre- 
quency text against stripped code 

NC7: percentage selector whose special relay box 
allowed round-robin repeat searches and selected 
them on a percentage of coincidence basis 

NC8: automatic circuit changer, which allowed 
automatic switching of alpha or numeric data among 
as many as twenty-five plugboards and the rearrange- 
ment columnar data 

NC9: only a prototype of this special substitution 
punch was built. 

NC10 and NC11: typewriter-like near off-the-shelf 

NC12: replaced the NC3 

NC13; converted IBM cards to and from micro- 
film, if it worked; the conversion from microfilm to 
cards was a true innovation at the time 

58. (U) Interview with Fred Parker, and his award- 
winning article, "The Unsolved Messages of Pearl 
Harbor," Cryptologia 13(1991): 295. 


Page 81 


59- ffS#MJ NSA CCH Series XII Z.John A. 
Skinner, "The CXCO Story," NSA Technical Journal, 
VX1 (Fall 1971), 21-37. 

60. (U) IBM would offer similar equipment to 
commercial users after the war. For a list of 
Letterwriter CXCO equipment available from the 
newly named Justo-writer division of IBM in 1947, see 
Hagley Museum and library', Accession 2015, ERA 
Materials, "Seminar Meeting, Tuesday, March 11, 

61. (U) Private Paper on NSA Machinery, 1985. 
NSA RAM File, CNO, U. S. Naval Communications, 
CITP TS Machine Descriptions, Washington, D.C., 
circa 1945- "Letterwriter." NSA RAM File, CNO, U.S. 
Naval Communications, CITP, "Machine 
Comparisons," June 1946. 

62. (U) The name was frequently spelled as 

63. (U) NSA RAM File, CNO, U.S. Naval 
Communications, CITS Technical Paper TS-48, 
"Machine Comparisons," June 1946. 



Chapter 4 
(U) Meeting the Crisis: Ultra and the Bombe 

(U) Looking Ahead - Ultra Saves RAM and 
OP-20-G Creates a Science Company 

(U) The history ofOP-20-G's cryptanalytic 
machine program would have been very different 
if Britain had had the power to read the German 
U-boat messages during 1942. On the chance that 
its men could beat the British to a reentry into the 
U-boat Enigma, OP-20-G was granted its long- 
sought Rapid Machine program and its own fac- 
tory and workforce. But the establishment of 
what became known as the Naval Computing 
Machine Laboratory came at a price. Because the 
American navy had not attended to the Enigma 
and because Hooper and Wenger's pleas for 
machine development in the 1930s were not fully 
heeded, OP-20-G had to defer its attempts to cre- 
ate advanced electronic computers for pure crypt- 
analysis. To solve the "E" problem, the machine 
group spent most of its first year and one-half, 
and much of the next two years, coaxing electro- 
mechanical components into doing things never 
before expected of them. The conglomeration of 
electrical, mechanical and electronic parts called 
Bombes turned Engstrom's men away from solv- 
ing the fundamental problems associated with 
Bush's designs, and away from an exploration of 
the possibilities of a general-purpose electronic 
machine. * 

(V)The "E" Machine 

(U) The Bombe was the example of the need 
for a technological retreat to deal with a crypto- 
logic emergency. Despite Germany's destruction 
during World War I and the crippling burdens 
imposed on it under the peace, it built a strong 
codemaking capability during the 1920s and 
1930s. At the center was the Enigma encryption 

machine, the workhorse of its military communi- 
cations networks. 

(U) The Enigma was a typewriter-size device 
that could be used in the field as well as in an 
office. It was electromechanical and used batter- 
ies to provide the electric current which passed 
through a series of shifting transposition rotors 
(commutators) to yield an extremely long encryp- 
tion cycle. Physically, Enigma consisted of a key- 
board to enter letters, a cascade of moving wheels 
that scrambled their inputs, a reversing wheel 
that sent the electrical impulses back through the 
wheels, a plugboard that further mixed the let- 
ters, and a series of lamps that showed the final 
result of the encryption. 

(U) The Germans felt safe because they calcu- 
lated that even if the wiring of the code wheels 
were known, it would take impossibly long for an 
enemy to identify the particular "key" settings of 
a message. In its early configuration, with just 
three of five available code wheels being used and 
no plugboard, Enigma had over one million pos- 
sible settings. 2 

(U) When the plugboard was added to the 
military versions, the Germans felt even more 
confident. The possible combinations jumped 
into the range of two hundred million million mil- 
lion! That made intellectually blind attacks on the 
machine an impossibility. 3 

(U) One World War II cryptanalyst explained 
why there was a critical need for revolutionary 
methods, and machines which could reduce the 
number of possible "E" settings that had to be 
examined, by stating: 4 "If every man, woman and 
child in the British Isles were given an Enigma 
machine, they would have to try 3,000,000 pos- 


Page 83 


sibilities on each starting position on each wheel 
order and would work their whole life to break 
one key." 

(U) The Germans constantly changed the "E" 
to make it stronger. They enhanced the system by 
increasing the number of wheels to choose from. 
By the end of the war, German codemen could 
choose from as many as nine main scrambling 
wheels when selecting a setup for their machines. 
In 1942 the Atlantic U-boat system added a 
fourth wheel inside its machines. Later in the war, 
the Germans made the machine more robust 
when they changed the plugboard and attached 
the Uhr box. The Uhr added another level of 
complexity by eliminating some of the cryptolog- 
ic weaknesses of the plugboards. 

SCMtUStit M 



(U) Qiigma 

(U) Perhaps most frightening to the Allies was 
the "pluggable reflector" which appeared on some 
German air force and army networks near the end 
of the war. It created more combinations to hunt 
through than had the introduction of a fourth 
wheel. 3 

(U) The Germans were particularly sensitive 
to a weakness in all encryption systems, the vul- 
nerability of internal indicators. Indicators were 
the brief instructions in each transmission that 
told a recipient how to set the remaining compo- 
nents of his Enigma, the ones not specified in a 
network's instruction books. Unlike the keys 
specified for all users in a network, indicators 
were selected by operators and changed with each 
message. Unfortunately for the Germans, they 
found no way to prevent their enemies 
from using those indicators to pene- 
trate some of the "E" networks. The 
exploitation of the indicators was one of 
the most important ways that Poland's 
cryptanalysts entered the Enigma sys- 
tems in the early 1930s. 

(U) The Poles were helped by stolen 
documents and used many more 
approaches than the attack based on 
the indicators. 6 

(U) Only a Few Were Able and 
Willing to Tackle "E" 

(U) Poland created what Stanford 
Hooper and William Friedman yearned 
to have in America: an office devoted to 
pure cryptanalysis. Poland's codebreak- 
ing bureau was able to recruit several 
bright young mathematicians who, as 
early as 1930, began to apply group the- 
ory and other advanced mathematical 
and statistical techniques to the 
German Enigma system. With the help 
of stolen documents provided by the 
French, the Poles began to understand 
and then penetrate the Enigma. They 


Page 84 

TOP Ct:cnCT»0OMI»T//nCL TO UOA, auo, eAH, UB I l AU P WB3BC1 


were reading many German systems by the mid- 
19308. By supplementing their mathematical 
analyses with the weaknesses of some of the oper- 
ational uses of the Enigma, such as repeating the 
indicators for a message or picking keys in a non- 
random way, the Poles were able to avoid using 
brute force searches that tested every possible "E" 
setting. They even learned how to avoid using 
data- heavy statistical analysis. A significant and 
fundamental discovery by the Poles was that the 
forbidding and seemingly impregnable plugboard 
was irrelevant in some cryptologic contexts. The 
discovery about the plugboard reduced the num- 
ber of tests needed to identify an Enigma's setup 
by millions. 

(U) The Poles Automate Cryptanalysis in 
Their Special Way 

(U) The Polish group also called upon 
automation in the early 1930s. Much work and 
genius went into the invention of an electro- 
mechanical machine, the Cyclometer, which 
automatically generated all the patterns produced 
by various Enigma settings. The Cyclometer was 
not a statistical machine or a device that could 
lead to a modern computer, however. It was an 
electromechanical rig that produced a card cata- 
log so analysts, in just a few minutes, could go 
from the indicators in a message to the Enigma 

(U) In 1938, to meet a change in the way the 
Enigma's settings were communicated, the Poles 
invented their version of an electromechanical 
automaton, the Bomba. 7 The Bomba was a set of 
linked Enigma machines that tested for the letter 
cycles produced by the setting indicators in 
Enigma messages. 

tS^JOne explanation for the use of the strange 
name, Bomba, is based on the mechanical crude- 
ness of the first Polish machines. To save precious 
construction time and parts, when the Bomba 
found a hit, a weight on the side of the machine 

was dislodged and dropped to the concrete floor 
with a very loud "bang." 8 

(U) The special-purpose Bomba was based 
upon a negative logic and used a special "crib" 
composed of the message indicators. The 
Bomba's goal was to eliminate the wheel orders 
and wheel positions that could not have produced 
the letter-to-letter cycles in an indicator. 

(U) The Poles had constructed six of the 
Bombes, one for each possible wheel order. That 
was adequate when the Enigmas came with only 
three encryption wheels to choose from. But just 
as the first Bombas were put into operation, the 
Poles had to face an increase in the number of 
Enigma wheels, then an alteration in the use of 
the "E" plugboard. Those changes demanded ten 
times the number of Bombas for a timely search. 
The Poles were too exhausted to produce so many 
additional machines. Their attempt to reenter 
Enigma through new statistical and hand meth- 
ods was frustrated by a lack of manpower and 
time. 9 

(U) Keeping the Bomba Secret for Too Long 

(U) When the invasion of Poland seemed 
imminent, and when the Warsaw team could not 
sustain its automation efforts, the Poles started to 
pass their secrets to their friends. 10 It was not 
until late summer 1939, when the Poles had to 
have help in producing more of their vital overlay 
sheets and Bombas, that Britain and France were 
informed of how the Polish men had been able to 
read Enigma messages. 

C FS//SF) The British representatives were 
grateful for the information, but they also were 
very upset that they had not been told the secrets 
when they had met with the Poles in February 
1939. Their anger almost led to a break between 
the two sets ofcodebreakers. One high-level 
British codeman, not realizing the Poles under- 
stood English, vented his frustration by berating 
them while in their limousine. Fortunately, the 


Page 85 


diplomatic skills of his countrymen calmed the 

(U) Despite the affront, the Poles gave the 
Englishmen copies of the Enigmas they had 
reverse-engineered and told them of the many 
ways to identify the various German Enigma 
radio networks. But on the eve of the invasion of 
Poland, the Germans made several more shatter- 
ing changes in their Enigma systems, which made 
Britain's task nearly impossible. After France was 
overrun, Britain was left with the responsibility 
for making a new beginning against the 
Enigma. 12 To exploit both her own previous work 
and the gifts of the Poles, Britain expanded its 
Government Code and Cypher School (GC&CS) 
and established the now famous Bletchley Park. 

(U) A Fresh Start against U E" 

(U) Despite the belief among many British 
influentials that the German code and cipher sys- 
tems, especially their naval ones, would never be 
broken, Britain made a significant commitment 
to cryptanalysis. ^ 

(U) Teams of brilliant men and women were 
recruited from the universities to work on the var- 
ious Axis systems. Alan Turing was only one of 
the Bletchley wonders who were recognized 
experts in mathematics and logic. 14 Under 
intense pressure, by 1940 he and others at 
GC&CS began to create the many invaluable tech- 
niques and electrical devices that eventually gave 
birth to the Ultra Secret. For example, the first 
Bletchley version of its Bombe was in operation in 
early i940.The next important configuration, 
with the ingenious diagonal board for the critical- 
ly needed simultaneous testing of plugboard set- 
tings, was running by August. 15 That Turing- 
Welchman Bombe of 1940 was a cousin, but a 
very distant one, of the Polish Bomba. 

(U) Turing explored many varieties of possi- 
ble solutions, hoping to find one that would with- 
stand changes in"E" and its usage. Although 

actively seeking pure methods to attack the 
Enigma, Turing eventually had to accept the use 
of a dependent and near brute force approach.' 6 

(U) At first he thought he had discovered a 
relatively pure method. In 1939 when he went to 
the naval section at Bletchley Park, Hut 8, Turing 
sought a robust and universal means of attack. 
After learning as much as possible about Enigma, 
he called on his knowledge of statistics and prob- 
ability. He arrived at a method quite like the one 
Wenger and Bush had chosen for the 
Comparator, the Index of Coincidence, The name 
given to his cluster of statistical methods was 

■fPS-J-Turmg focused his statistical powers on 
the German naval systems because they had been 
the most intransigent. He thought that if his 
"Banburismus" methods proved of worth against 
them, they could be generalized to all cipher 
machine systems.' 7 The general logic of 
Banburismus was, like the IC, based on the statis- 
tical characteristics of language. The goal of both 
approaches was also the same: to identify mes- 
sages that had been enciphered with the same 
"key" or machine setting. Once such a "depth" 
had been pinpointed, the machine setting could 
be found and the cipher messages turned into 
readable text. 

cW)-More than the logic and goals were simi- 
lar. The techniques were essentially the same. 
Two messages that were thought to have been 
produced by the same key (as suggested by such 
evidence as the same callsigns and indicators), 
were placed one above the other. Coincidences 
were then counted and evaluated against the 
number expected by chance. The counting was 
repeated for each ofthe offsets. Turing even 
mechanized the process through the use of over- 
lay sheets. Holes were punched in the sheets to 
represent the text. When the two sheets were 
superimposed, the coincident holes were very 
easy to identify. That allowed relatively unskilled 
labor to be used to tally the results. The overlap 

Page 86 


TO P 5E&H E T//C0 MI HT//REL TO UCA, AUG, CAM, rtBP flMn N7 I »Y1 

method could also be used to exploit the informa- 
tion that could be drawn out of the indicators in 
the messages. 

CfiS) Turing went further with his ideas. He 
developed his IC-like approach into an elegant 
predictive system. His "bans" were statistical esti- 
mates of how likely it would be that two messages 
would prove to be of value in identifying the 
cipher keys. Such estimates allowed the cryptan- 
alysts to make rational decisions about allocating 
their very precious time. They could concentrate 
their skills on the messages most likely to yield 

(■'iS} Banburismus was Britain's initial 
method of attack on the German naval Enigma 
and continued as its most powerful tool until 
1942. Helped by other techniques, such as 
"scritching," 18 it was the way Hut 8 identified two 
of the three wheels used in an Enigma setting. 
Once the wheels were specified, the analysts 
could attack the other parts of the Enigma "key." 

CSS^ Unfortunately, while Turing's 
Banburismus was an advance on the state of 
cryptanalytic art, it was not strong enough to be a 
timeless and independent conqueror of Enigma. 
Its target of the early 1940s, the naval Enigma, 
was too rugged. The success of Banburismus 
depended, despite Turing's hopes, on knowing 
the contents of the very special "E" instruction 
sheets the German Navy used. From 1939 to the 
end of 1942, when Banburismus was no longer 
employed against the naval systems, the British 
had to capture or, with a great expenditure of 
manpower, reconstruct bigram and other com- 
plex tables that were used to superencrypt the 
naval "E" indicators. Banburismus went blind 
several times when the Germans changed the 

(TS//8I) As used against the naval systems, 
Banburismus was also dependent upon having a 
very large number of messages in "depth." As 
many as 300 messages might be needed to allow 

the identification of the Enigma wheels. 19 Even 
with enough messages, Banburismus needed 
more help. 

CTS//SI) Banburismus was elegant, but it was 
not self-sufficient. Wheel wiring had to be known, 
and "cribs" and much hand testing were required 
to identify the plugboard connections and wheel 
turnover points. 2 ° A very large investment had to 
be made in compiling a catalog of all the possible 
enciphernents of the German word for "first." It 
was needed to supplement the statistical analysis 
with primitive cribbing. That "Eins" catalog 
drained the resources of Bletchley Park, but it did 
not prove as timesaving as hoped. Moving from 
the suspected location of the crib word "Eins" to 
the catalog entry for its possible setting and then 
testing to see if the key had truly been found were 
too demanding. 21 

fK*Hn near desperation, Turing turned to a 
full-blown crib approach. He relegated 
Banburismus to being an adjunct to his version of 
the Bomba. By 1940 he had designed a machine 
that was to be a high-speed, automated, and near 
universal "catalog." His device would take any 
long crib and test it against all possible wheel set- 
tings of an Enigma and do it within minutes. 

(U) Given the technology available to him, 
this rather crude method seemed the only alter- 
native. Thus, while he explored the application of 
other statistical methods to the "E" problem, he 
sketched out a new Bombe. 22 It used some of the 
ideas of the Polish Bomba, but the British Bombe 
and its logic were special. 

(U) Turing's Bombe was an electromechanical 
analog of the Enigma. 23 It was based on identify- 
ing logical contradictions as represented by flows 
of electricity. Its banks of interconnected high- 
speed "E" wheels spun until they found a setting 
that might have produced the crib setup on the 
machine. Like the Bomba, it needed to search 
through all the wheel settings, and it accepted the 


Page 87 

TOP OCOrcCTOOOMIHT/mCL TO I K A, A I J? , CAM, GRff fthlfl N7 I fOH 

consequences of relying upon a special-purpose 

(U) Turing's Bombe needed a great deal of 
prior information about German networks and 
their keys. His crib attack was premised on know- 
ing the wiring and turnover pattern of each "E" 
wheel, and it needed insights into the plugboard 
and other settings of each Enigma net. Turing 
knew his Bombe could have been made as useless 
as the Polish machine if the Germans significant- 
ly increased the number of letters changed by the 
"E's* plugboard, if they stopped using stereotyped 
phrases at the beginning of their messages, or if 
they ended the practice of sending "E" messages 
on simpler cipher systems. He also knew that his 
machines would be expensive and that their con- 
struction would perhaps ask too much of Britain's 

(TO) Turing faced some stiff opposition when 
he requested that a program be funded. Although 
he explained that Banburismus would reduce the 
number of wheel combinations that had to be 
tested from more than 300 to fewer than thirty, 
thus calling for only a dozen Bombes, administra- 
tors at Bletchley had serious concerns. 24 They did 
not want to waste money and time on the con- 
struction of a machine that had to correctly scan 
hundreds of circuits within a fraction of a second 
They knew the dangers in trying to construct a 
reliable machine that was to have ten miles of 
wire, a million soldered connections, and a clutter 
of mechanical parts. 

(U) More than the machinery was at risk. To 
find the right kind of cribs for the Bombes called 
for the creation of a new and large group of ana- 
lysts to constantly mine German intercepts for 
new leads. 25 

(U) Turing made his machine as universal as 
possible. Although it followed the logic of point- 
ing to the wheel settings that could not be elimi- 
nated, it tested the settings against letter loops 
from within relatively long plaintext phrases 

(cribs) in messages. Relying upon words within 
messages rather than indicators guaranteed a 
longer life to his Bombes and promised fewer 
false drops. 

(U) A bit of luck made the Bombe even 
stronger. Before Turing had finished his design 
for a machine to attack the three-wheel Enigmas, 
a young mathematician appeared at Bletchley 
Park whose insight multiplied the Bombe's abili- 
ties. Gordon Welchman's suggestion for the 
"diagonal board" allowed an instantaneous test 
for the influence ofthe plugboard setting and 
allowed the effective use of relatively "weak" 
cribs, ones without long letter loops. 26 

(U) Analog and Parallel May Be Fast, But ... 

(U) Although electronics was tempting and 
although men like Turing knew that digital pro- 
cessing would become the basis for modern com- 
puters, a large number of machines had to be put 
in operation in weeks, not years. Britain needed 
working machines immediately. In late 1939 
GC&CS's managers had to turn to someone who 
could produce immediate and sure-fire technical 
results. They found the right man: "Doc" Harold 
Keen, the head engineer at Britain's version of 
IBM, the British Tabulating Machine Company. 
Keen built a prototype in a few weeks and was 
able to begin sending some operational Bombes 
to GC&CS in a few months. One reason for his fast 
work was the use of standard, tried-and-true 
parts and analog logic. 

(U) To match Turing's logic, Keen designed 
new five-inch commutator drums that were hard 
rubber and metal contact imitations of a double 
Enigma wheel. 27 The drums were arranged in 
banks of three, each being a double analog of an 
Enigma scrambler unit. One wheel in each bank 
was run continuously, another moved after a full 
revolution ofthe first, and the last stepped after 
the other two had completed their cycles. 

Page 88 



(U) As the Bombe's wheels spun over the 
commutator connections, they created instanta- 
neous multiple electrical pathways through the 
other banks. Then the electrical charges went to 
the relays that matched the flows against the crib. 
At the same time, they surged through the 
Welchman diagonal board to test the assump- 
tions about the setup of the "E" plugboard. 

(U) In the first models a great deal was 
expected of the operators. To identify the wheel 
positions when a hit was encountered, they had to 
touch the relays. Eventually, a small printer was 
attached to the machines. 

(U) Keen's engineering task was made more 
difficult by the need to test for another type of log- 
ical impossibility. To make the test, he had to pass 
the output of the wheels through the diagonal 
board. The diagonal board was a twenty-six by 
twenty-six matrix of resistors that instantaneous- 
ly sensed inconsistencies such as two different 
input letters being enciphered into the same out- 
put letter. 

(U) To test all possible wheel combinations 
against just one crib for a three-wheel Enigma 
called for at least sixty machines. The Germans 
ran dozens of systems, and only hundreds of 
Bombes could have provided unaided coverage of 
them all. Turing's method of reducing the num- 
ber of wheel combinations that had to be tested 
was soon overwhelmed by improvements in the 
German systems and by the proliferation of 
encryption networks. 

(U) Unfortunately, BTMC faced too many 
shortages of men and materials to keep up the 
early pace of production. Keen was able to send 
less than a machine a month to Turing during 
1940 and 1941. In early 1942, the record was not 
much better. Bletchley had only sixteen bombes 
and production had slowed. 28 To produce two of 
the Bombes a month stretched Britain's produc- 
tive capacity. Bletchley Park was unable to build 
up enough of an inventory of them to seriously 

challenge any Enigma system until the end of 
1942 and in the first days of 1943 GC&CS still had 
fewer than fifty machines. 

(U) But it was not until mid- 1942 that 
Britain's leaders decided to commit massive 
resources to the Bombe program. Only when they 
seemed essential to victory in Africa and to the 
safety of the Middle Eastern oil supplies was 
"Doc" Keen given new factories and a large work- 
force. That allowed BTMC to produce some 200 
three-wheel Bombes by the end of the war. 29 

(U) Although Britain did not have much of an 
Ultra Secret in the critical months of 1940-41, she 
wanted to keep what she had to herself. Britain's 
codebreakers feared revealing their methods, 
even to the Americans, whose military aid had 
become essential to their nation's survival. 30 They 
had a not unwarranted fear that sharing with 
America would lead to breaches of security and 
the demise of Ultra. 31 

(U) The World War II relationship between 
the British and American cryptanalysts began in 
confusion and mistrust. It took several years to 
reach workable accords, and the formal, long- 
lasting agreements came after, not during, the 

(U) The trust that became the foundation of 
the Cold War cooperation between the two 
nations did not come easily. There were critical 
months in late 1942 and mid-1943 when it 
appeared that what had been achieved since 1940 
would be lost. The combination of British reluc- 
tance, America's divided armed services, misun- 
derstood agreements, and lost messages almost 
led to an end to the joint intelligence program. 

(U) Ask and Then Not Receive 

■4$} Britain's leaders had begun making over- 
tures about sharing "scientific" information as 
early as February 1940. When the suggestions 
reached the American Army and Navy codebreak- 

TOP S C eW E TOe O M I N T /mEL T O U3A, AU3, CA N , QDR A N D N 2L//X 1 

Page 89 


ers, they did not reject the possibility of some sort 
of exchange. But Safford and his superiors in 
Naval Communications soon cooled to the idea, 
leaving William Friedman as the only advocate 
among the cryptanalysts. 32 

(U) Friedman could not deal face-to-face with 
the British; frustrations grew and there was a 
break in the negotiations. There may have been 
vague promises offull cryptologic cooperation 
between Roosevelt and Churchill in mid-1940, 
but they did not lead to any significant exchanges 
among the codebreakers. 33 

(U) Then after Britain began to send the 
United States Navy information on the disposi- 
tion of German forces, it appeared that an agree- 
ment about an exchange of secrets was immi- 
nent 34 

(U) In September 1940 William Friedman 
drew up a detailed plan for cooperation between 
the two nations only to encounter a wary 
American navy that again blocked its implemen- 

-fj9}-But the navy did not have its way. The War 
Departments representatives made agreements 
for «fcll» exchange of information in December 
and exerted enough political pressure to sweep 
away the navy's objections. 35 OP-20-G was given 
orders to cooperate. It was told to select two men 
to join an army and FBI team that was to sail to 

-(S}-The navy was unhappy about being forced 
to exchange its secrets in spring 1941. Then when 
the Americans concluded that what Britain was 
willing to show the Americans was much less 
than expected, some in the American cryptanalyt- 
ic community became furious over the balance 
between what America gave to GC&CS and what 
it got in return. J 

(U) Gave All and Got.,. 

(U) The delegation of American cryptanalysts 
from OP-20-G, the army's SIS, and the FBI had 
sailed for England in late January. They handed 
over two extremely valuable analogs of the Purple 
machine for the Japanese diplomatic ciphers, two 
copies of another Japanese enciphering machine, 
and all the other keys to the top secret Magic sys- 
tem. In addition, all that the United States had on 
major Japanese attache, navy, and consular codes 
was surrendered. As with the Purple machines, 
giving those paper copies of the codes to Britain 
meant fewer were available to America's own 
codebreakers. 37 The American generosity did not 
end there, however. Britain was promised a con- 
tinuous flow of cryptologic information, including 
the American Coast Guard's methods of tapping 
the German clandestine systems." 38 

(U) In return, GC&CS opened its doors and 
made the American visitors feel quite welcome. 
But it gave them very little of real value, at least 
about "E." 

£ ES//SI) At first the Americans thought the 
British were completely open. Although the 
Americans were told to pose as Canadians, they 
felt that few restrictions had been imposed on 
them. They accepted the order against taking any 
notes on what they were shown and took having 
to sign a binding security oath as reasonable. 

-fS^-They felt they had been told all about the 
British attack against Italian, South American, 
and Russian systems. 39 GC&CS also shared its 
work on Japanese naval systems. And the 
Americans were shown the Bombe. The navy's 
men were given a paper version of an Enigma, 
were handed a copy of a few days' worth of old 
keys, and given a part of a "short" catalog. 

(U) The army's representatives were given 
similar information, and they and the navy men 
were informed of the earlier successes against the 

Page 90 



German air force's "E" and the system Germany 
had used during the Norwegian campaign. 40 

(U) What they did not get was what many had 
thought the trip across the Atlantic was really for, 
the cryptanalytic keys to the naval Enigma. The 
Americans did not obtain an Enigma machine or 
enough cryptanalytic information to allow the 
United States to break into the submarine "E" on 
its own, 


(U) What Happened After 

(U) Although the American visitors to 
Bletchley Park may have left too soon to be told of 
the successes of mid- 1941, the British could have 
been much more open than they were during the 
remainder of the year. 

fGS* Despite GC&CS's proclamation that full 
cooperation was in force, the American navy men 
had not been given an adequate explanation of 
the logic of the Bombe during their visit, were 
probably not indoctrinated into Banburismus, 
and were told little about the art of obtaining 
cribs for the naval Enigma attack. 42 The FBI and 
army representatives were also not told all. 

TFS-)-A U. S. Navy historical report on the "E" 
problem stated: 43 "Prior to the outbreak of the 
war with Germany, the nature of the German 
machine employed by the Atlantic U-boats was 
known in that the British had supplied to this 
Division diagrams of the wiring and the wheels of 
the device, together with a description of the way 
in which it moved. Beyond this and some few 
examples of plain text, nothing was known as to 
the usage of neither the device nor the method in 
which the keys could be recovered. It was then 
known that the British had conducted a success- 
ful attack, but the details of it were unavailable to 
the American Navy, due to the reluctance of the 
British to discuss the same." 

(T3//0I) The desire of the British to safeguard 
their secret powers was reflected by their failure 

to communicate the news or details of their 
achievements to the Americans after March 1941. 
Little cryptanalytic information crossed the 
Atlantic although some vigorous protest came 
from OP-20-G beginning in July 1941. 44 Even a 
visit to America in late summer by a very impor- 
tant British cryptanalyst did not lead to the 
Americans being told of the ways to attack "E." 45 

(U) Trust Builds Very Slowly 

(U) For a year after the American delegation 
left England, there were few direct contacts 
between the two nations' cryptanalysts. There 
were some negotiating sessions about the range 
and degree of cooperation, but during the 
remainder of 1941 it seemed to many Americans 
that Britain became less, not more, willing to 
yield its growing pool of Enigma secrets. 46 

(TS) The situation became quite tense by 
November. OP-20-G's men convinced the 
Director of Naval Communications to send a very 
strong protest to England. He told GC&CS he 
thought that the earlier agreement was not being 
fulfilled and demanded an immediate flow of 
cryptanalytic information. The British responded 
with true diplomacy explaining that all that had 
been promised had been sent to the United States 
and that it would be impossible for them to send 
everything that the American navy "might" want. 
It was better, they said, for the Americans to 
request specific information. Then if the British 
judged it was really of import to the United 
States, it would be sent. A quite similar message 
was forwarded to the American army. Although 
England began to send information on German 
diplomatic systems to Friedman's group, England 
continued to keep its "E" methods a secret. 47 

(U) Agreements and Agreements and 
Agreements, But.... 

(T0//0I) OP-20-G was, of course, very unhap- 
py about being required to ask for specifics; they 
did not know enough to compile a list. But the 


Page 91 


tension was relieved by some end-of-year 
exchanges. When Laurance Safford finally 
received some replies to his earlier inquiries, 
when Britain hinted they would soon send a copy 
of their "machine" and a technician to America, 
and when GC&CS apologized for losing some 
American letters within its bureaucracy, "G" 
sensed it was going to be made a full cryptologic 
partner. 48 

(U) Under some prodding from the United 
States, additional agreements were made in early 
and mid-1942, ones that began to move the two 
nations toward a level of unprecedented coopera- 
tion. Then the sweeping October 1942 accords 
eased some tensions raised by an American navy 
threat to go its own way on Enigma. After that, 
the BRUSA pact of May 1943 was a major step 
toward openness with the army. 49 

(U) But it was not until the UKUSA agreement 
of 1946 that the two nations forged that unique 
relationship of trust that was maintained 
throughout the Cold War. 50 

(U) There were more than a few frictions on 
the road to BRUSA and UKUSA; during 1942 and 
1943 the British were slow to reveal all about 
Ultra, especially during 1942. The American 
cryptanalysts had interpreted the agreements of 
1941 to mean that Britain was to share all and that 
America was to become a full partner in Ultra. 
Although negotiating separately, both "G" and the 
SIS concluded the same thing. They expected that 
Britain would give them all they needed, if they 
wished to read "E* systems. A copy of the British 
Bombe designed for the older three-wheel 
machines was expected by "G" before the summer 
of 1942, and some Americans thought that all 
British information on new Bombes for the naval 
four-wheel Enigma would be immediately sent 
across the Atlantic. The British did not seem to 

(U) Going Separate Ways? 

(T0//0I) When "G's* frantic February 1942 
plea for help against Enigma in the Atlantic did 
not get a response, tempers flared again. OP-20- 
G's new commander, Captain Redman, under 
intense pressure from Admiral King to do some- 
thing about the Atlantic submarine crisis, con- 
cluded that GC&CS had been giving America the 
"runaround." It did not take him long to secure 
permission to begin another and very determined 
series of negotiations with the British. With 
Joseph Wenger at his side, Redman began to 
make it very clear that unless cooperation began 
immediately, the American navy would go its own 
way despite any of the danger that two uncoordi- 
nated attacks on Enigma might pose. 

^JS^ When the Americans came to realize that 
the U-boat commanders had made a radical 
change in their "E" systems and that the British 
claims of imminent reentry were far from true, 
they began to take action. Sl 

f TO) ■Britain sensed there was a crisis and 
decided to send one of its most important code- 
breakers to America. He was told to calm the 
Americans without, however, giving them 
GC&CS's great secrets. When Colonel John 
Hitman arrived in April 1942, he found it impos- 
sible to agree with official British policy. The 
American navy's codemen were so adamant and 
their threats to go their own way were so credible 
that Hitman advised England that it must yield. 52 

('ESKThe final breakthrough seemed to come 
in May 1942. Promises came from England that a 
Bombe and a technician would be in America 
before autumn and that men from "G" would be 
invited to Bletchley Park. 53 There was some hope 
that the American army's cryptanalysts might be 
allowed into the "E" circle. GC&CS signaled that it 
would accept some help from Leo Rosen, SIS's 
electronics expert. 54 

Page 92 

top ocoRCT/fooMiNTy/RCL TO USA, aus, can, obr a i ib mzu/xi 

top qi:gri:t//co mi nt//rcl to uoa, auo, can, odr a n d nzu/x i 

TK>}- But the May promises did not end 
American suspicions, nor did they lead "G" to 
trust the solution of"E" to the British. Adding to 
the tensions, there was disagreement over the 
details of when and how "G's" experts were to be 
allowed back into Bletchley Park. Then, when a 
copy of the latest English Bombe was not shipped 
to America on schedule, when it took six months 
of requests to obtain promised blueprints, when 
Britain kept insisting that the United States work 
only on Japanese problems, and when it was clear 
that England did not want to give anything to the 
SIS, some American codebreakers again became 
very skeptical of British intentions. 

T^S) The suspicions led to action. Despite the 
recent accords, in August 1942 OP-20-G felt it 
had to protest about the failure of the British to 
keep their promises. Also, the relations between 
SIS and GC&CS became very tense. The army- 
British relationship became so strained that 
protests reached the White House. 55 

(U) America without an Ultra 

(U) At its entry into the war, OP-20-G had 
only the most rudimentary knowledge of the 
Enigma and was not at all sure about the contours 
of the new U-boat system. 56 Some disgruntled 
American officers blamed Britain's unwillingness 
to share, but the reasons why American cryptan- 
alysts were helpless lay in America, not Europe. 57 

(U) Since the turn of the century, America's 
strategic planners had seen Japan as the enemy. 
Some in the American military did worry about 
Germany, but it seemed beyond imagination that 
France and Britain would be unable to contain 
her on the continent. Few thought they would fail 
to block Germany's navy and air force from mak- 
ing the Atlantic unsafe for America. The concen- 
tration on Japan led to another dangerous 
assumption: No matter what the enemy did, the 
United States would have the time to prepare 
itself for war. 

(U) Those assumptions were accompanied by 
ones about the nation's economy. Fundamental 
was that American industry would automatically 
provide any great technological advances needed 
by the military. It is no wonder that the calls by 
men such as Hooper and Bowen for ongoing 
research and Bush's drive to establish govern- 
ment-sponsored science remained largely unan- 
swered. 58 

(U) Despite a lack of resources, by spring 1941 
American naval ships were involved in dangerous 
scrapes with German U-boats in the Atlantic. By 
autumn the Americans were ordered to escort 
England-bound convoys. 

(U) OP-20-G was as unprepared as the rest of 
the navy. To fulfill its obligations in the Atlantic, 
it expanded its interception net. To please 
England, it put most of its men on tactical analy- 
sis rather than codebreaking. At the same time, 
the navy's cryptanalytic ally, the Coast Guard, 
launched an attack on German clandestine mes- 
sages. But the spy messages and the bits and 
pieces from some cracks in the German diplo- 
matic systems yielded little about the German 
navy. OP-20-G had no effective Atlantic crypto- 
logic power, and the navy had to rely upon British 
supplied intelligence. 59 

(U) When the U-boat command changed its 
Enigma and Hitler unleashed his American war, 
OP-20-G's cryptanalytic weakness became intol- 
erable. When it was realized that Britain was 
closed out of the U-boats' new M4 Enigma Shark 
system and as Britain seemed more interested in 
the German army and air force systems, OP-20-G 
decided to find its own way to penetrate Enigma. 

(U) An American Ultra, Perhaps 

(T0//0I) In spring 1942 the American navy 
was ordered to start forging its own "E" capabili- 
ty. Despite the crisis in the Pacific and the old 
hopes of building general-purpose computers, 
Howard's men and those in "M" were ordered to 


Page 93 


focus on the Atlantic Shark problem and to pro- 
duce an immediate solution. 60 

(U) When the decision was made to create its 
own "E" solution, OP-20-G was very short-hand- 
ed and had to turn much of the work over to 
Engstrom's group of college men in the "M" sec- 
tion. 6l They began with few tools and the burden 
of Britain's fears of an independent American 
Ultra capability. 

(V) Faster Than a Speeding Relay 

(U) Bletchley Park's very overworked men 
had let almost a year slip by without focusing sig- 
nificant resources on a Bombe for M4, the four- 
wheel U-boat Enigma. 62 The Germans* introduc- 
tion of additional encryption wheels an operator 
could choose from also presented a great chal- 
lenge. Changes in the related German codes, 
radio networks, and procedures compounded the 

(U) GC&CS called on the famous Wynn- 
Williams and asked him to explore the use of elec- 
tronics for a super-bombe. Williams spent many 
frustrating months trying to create an electronic 
Bombe, His efforts stretched into spring 1942 
with little more to show than a breadboard model 
of a primitive "E" wheel. While he was asked to 
make a fresh start on his ideas, GC&CS turned 
back to "Doc" Keen and BTMC. Immediately the 
pragmatic Keen rejected an electronic solution 
and began to give some thought to alternatives. 63 
He created the outlines of a new four-wheel 
Bombe but advised GC&CS that it might take 
more than a year to design and build the first 
model. 64 His prediction proved correct. Britain 
would not have the first of its very few tempera- 
mental electromechanical four-wheel Bombes 
until early summer 1943. 65 

(U) Great British Expectations 

(U) Meanwhile, Wynn- Williams continued to 
plod along with his ideas; by midyear he began to 

construct a prototype of his Bombe. To his disap- 
pointment, it had to rely as much on mechanics 
as on electronics. 

™(T0// 01) Williams had decided to build a 

complex attachment for the regular three-wheel 
British Bombes. His Cobra was to be a large box 
that was to contain his new high-speed electronic 
wheel and newly designed control and "hit" loca- 
tion devices. At first he bet that he could coax 
electronics to do the entire job. It took a long time 
to admit even partial defeat, but he had to back 
away from his original plan. He turned his Cobra 
into a combination of very high-speed commuta- 
tors (3,000 rpm) and electronic memory and 
control circuits. 

ffl)//SI) - The compromise did not lead to 
immediate success, however. Williams had asked 
a great deal of electronics and mechanics. The 
Cobra was planned to be exceptionally fast. In 
addition to the tube circuits and the new wheels, 
run time was to be shortened by recording "hits" 
without stopping the Bombe. All that was too 
much for Williams' small team. His first machine 
had to suffer the indignity of a thorough rework- 
ing at the end of 1943 before it had done any oper- 
ational work. Although a dozen of his new 
Bombes were eventually employed in England, 
they remained temperamental. 66 

(U) Great American Expectations 

(U) While Williams and Keen were rushing to 
find their technological answers to the M4, 
GC&CS learned of America's Ultra intentions. 
Frightened by what it discovered, in spring 1942 
it rushed a group of its leaders to the States hop- 
ing to reach an understanding that would protect 
its Ultra monopoly. 

(U) March 1942 saw Britain strike the first of 
a series of new bargains. It assured OP-20-G that 
Shark was about to be beaten, and it agreed to 
share more Enigma information with the navy. In 
exchange, it asked "G" to concentrate on the 

Page 94 



Japanese problems and let Britain manage 
European intelligence. The Americans desired 
cordial relations with the British, but they would 
not abandon Enigma. "G" agreed to cooperate but 
stood by its commitment to an American pro- 

(U) Despite two years of "understandings" 
with the British, OP-20-G launched into its own 
Enigma and Bombe programs without a true 
understanding of the Atlantic Enigma or the 
British Bombe. The historic OP-20-G directive of 
April 1942 gave a very incomplete view of 
Enigma, Shark, and the British Bombe. America's 
experts were able to outline only the workings of 
the older three-wheel plugboard version of 
Enigma, and they seemed uncertain about a key 
component, the reflector. Furthermore, OP-20- 
G's memorandum contained only the most gener- 
al ideas about the British Bombe's logic. 67 

(U) While waiting for the promised informa- 
tion from England, "G's" men were told to define 
a true American Bombe. Given all they did not 
know about Turing's and Welchman's methods 
and machines, the first plans for the American 
Bombe do not seem too bizarre. 68 

(TS//SI) OP-20-G made Howard Engstrom's 
young men responsible for the Bombe project. 
The closest "G" had to an Enigma expert, 
Lieutenant Commander R. B. Ely, was charged 
with devising the logic of the machine. After 
reviewing all the methods he knew to attack "E," 
he suggested that "G" might have to turn to a crib- 
based approach. 

(TS//SI) E ly, armed with only a few hints 
about how the British machine worked (gained 
through some test problems the British had pre- 
viously sent over), independently arrived at a 
primitive version of the architecture Turing had 
designed three years before. As soon as he was 
able to sketch the logic of his machine, he sought 
an engineer. Not unexpectedly, one of the young 
men from MIT was selected. 

(TS//SI) J ohn Howard was asked to solve the 
hardware and manufacturing problems. While 
Howard discussed possibilities with men such as 
Joe Desch of NCR, 69 Ely asked for help from oth- 
ers in Engstrom's section. He wanted assistance 
to check his ideas against cryptanalytic needs. 
And he wanted help finding, if it was possible, an 
architecture for a computer more universal than 
Turing's. Soon Ely's original ideas were reshaped. 

(TS//SI) The conception of what the 
American navy's Bombe would be was logically 
primitive but technically grandiose. In spring 
1942 "G's" men knew so little about the Enigma 
systems they did not include the important 
instantaneous stecker setting tester in their 
design, and they thought their machine could use 
permanently fixed "wheels." As important, they 
thought they would have to have either a separate 
Bombe for each of the wheel combinations and 
permutations or one truly giant machine. 

(TS//SI) Apparently, as late as early summer 
1942, Britain had not informed them of the "diag- 
onal board," the many methods its codebreakers 
had devised to reduce the number of wheel com- 
binations to be tested, nor how many different 
wheels the Germans made available with the M4. 
"G" did not know that the British had found it 
wise to leave tests for Enigma ring-settings and 
wheel starting-points off their Bombes. 70 

( TS//SI) "G's" technical visions were far from 
backwards, however. As soon as the idea of an 
American Bombe had emerged, electronics 
became the focus of attention. A breadboard 
model of a "wheel" was begun. Completing it 
would not be easy because the Enigma wheel was 
difficult to imitate, and constructing a universal 
one was a daunting task. Cloning one wheel with 
known wiring meant having twenty-six tubes 
connected to twenty-six others and having a rack 
of supplementary circuitry. A universal "wheel" 
needed 26 x 26 tubes and all the circuitry needed 
to switch them as needed to imitate any of the 
possible wiring connections. 

T*» iMTTiinn Tiinn in i.on »nc- r^n /jffrf ftHP M7 I ff v 1 

Page 95 


(T0//GB But Howard and the other navy 
engineers were so confident about electronics 
that as early as April 1942 Wenger was informing 
the British that American electronics might save 
their Bombe program. The faith in an electronic 
solution continued well into the summer, and 
some thought the American Bombe might turn 
out to be a single high-speed and complex tube 
machine that would do as much as or more than 
all of the British mechanical monsters. 71 

CRD//0I) Such a machine could save much 
time because it would also automatically reset the 
wheel(s) for each run and would never have to 
stop to record "hits." It was to have a high-speed 
system to photograph the diagonal board and a 
set of counters that would record the wheel posi- 
tions at each hit. 

C BS//S J Q Those time-savers were secondary, 
however. What was important was the speed of 
electronics. In early spring it was thought that the 
American Bombe could do its job if it performed 
10,000 tests a second. Faith in the machine's 
speed was necessary because it was going to be 
asked to do much more than Turing's Bombe. It 
would test for the some 300 possible wheel 
orders, the 440,000 [sic] stecker possibilities, 
and the ring settings. 72 

( TB//S t- ) Soon it was realized that even 
greater faith in electronics was required. When 
the run time for all that was recalculated, the need 
for much faster electronics was realized. Given 
the way the Americans were designing their 
Bombe at the time, one large one, a feasible 
machine called for circuits that could make mil- 
lions of tests per second. That kind of speed was 
far beyond the electronics of the 1940s, but there 
was no indication that frightened the Americans. 
They had such engineering optimism and knew 
they could command so many resources that, if 
one electronic machine would not do the job, per- 
haps 300 or so of their special electronic "dib- 
bers" could be built. 73 

(U) While Engstrom's men were exploring 
their options and while they waited for the 
expected flow of information about the Bombe 
and allied methods from England, W G" continued 
on with its fight to achieve Wenger's old dream of 
using pure techniques. OP-20-G's new college 
men intensified their search for advanced pure 
statistical and mathematical methods and 
machines. Hypo, Tessie and the Comparator were 
still seen as general- rather than special-purpose 

(U) But as summer arrived, the crew at "G" 
started to become angry and worried. Little help- 
ful information had come from England, and the 
two men OP-20-G planned to send to Bletchley 
Park were not scheduled to leave until July. 74 

fSf/SB-GC&CS was being more open with the 
Americans and informed them of the Fish sys- 
tems. But Britain's codemen still seemed to hold 
back on the Bombes and what was necessary to 
their success, the methods of finding surefire 
cribs. Although offering GC&CS full information 
on all the advanced high-speed cryptanalytic 
machines it was developing, 75 OP-20-G was made 
to wait for a reply to its specific requests and for a 
clear statement of British policy on cryptanalytic 
cooperation. 76 Especially frustrating were the 
delays in providing Bombe details. Requesting 
blueprints of Britain's "latest" machines in May, 
the Americans hoped their examination would 
prevent them from committing to an American 
Bombe that was inefficient or simply unworkable. 
The prints did not arrive as promised. 

(3S//SI) The Americans were under too much 
pressure to accept the continued stream of British 
"excuses." By the time Ely and Eachus were ready 
to depart for GC&CS, Joseph Wenger and his 
superiors became convinced that Britain would 
never finish the four-wheel Bombe they had 
promised to have working by mid- 1942. Worse, 
they thought that the British had not lived up to 
the agreements that had been made since 
America entered the war. Less than guarded 

Page 96 




words were used in some of the exchanges 
between "G" and England. 77 

(TS//SI) The British reacted to the American 
protests by sending more information, hoping 
that "G" would reverse its decision to build its 
own anti-Enigma capability. OP-20-G gradually 
learned more about Britain's cryptanalytic meth- 
ods, including those used to avoid testing all "E" 
wheel combinations. And when the two men from 
"G," Ely and Eachus, reached Bletchley Park, they 
began shipping home the detailed information 
the navy had sought for the last two years. 

- (S//SI) But the Americans remained very 
worried. They feared they would be unable to 
build their more universal machine or their own 
version of the English Bombe. Worse, OP-20-G's 
leaders deeply feared that even if they built such a 
Bombe, they would always remain dependent 
upon Britain for the necessary copies of captured 
"E" wheels, codebooks, and cribs. 78 

(U) Trying to Step Foruiard, Not Back 

(U) A few atOP-20-G were convinced that 
America could beat England's famed Wynn- 
Williams to a super-high-speed electronic 
machine, but others in the OP-20-G group were 
less sure of an independent American success. 

(U) Although Ely and Eachus were sending 
back important information, the navy continued 
to have to formally request much on the English 
Bombe and the emerging new solutions to the 
Enigma systems. More fundamental, by the end 
of the summer the Americans became concerned 
that Britain would never devote enough resources 
to the Atlantic U-boat problem. 79 There was some 
foundation for the American anxiety. 

(U) Britain's Own Version of Bush's 
Electronic Dreams 

(T0//0T) Tnlate spring 1942, as part of the 
reallocation of GC&CS resources, Wynn- Williams 

was asked to turn his Bombe work over to some- 
one else and to take on another job: devise a high- 
speed engine to crack the binary additive system 
of the Fish machines. He agreed, and while con- 
tinuing on with his electronic Bombe work he 
designed the first of the Robinson rapid analytic 
machines. 80 Very soon, the designs were turned 
into hardware, electronic hardware. 

(TOf/Ol) The first Robinson (Heath) was 
delivered in early 1943, well before any of the 
newest models of Bush's Comparators reached 
OP-20-G's headquarters. They were based on a 
statistical attack, not the type of crude crib-bash- 
ing of the Bombes. 81 Robinson used high-speed 
punched tapes, photoelectric readers, and some 
one hundred gas-filled tubes to keep track of 
results. 82 The Robinsons shared something else 
with Bush's machines, the very serious problem 
of keeping the tapes in alignment. There were dif- 
ferences, however. The Robinson's target was a 
binary additive system. That called for a different 
use of the tapes. One tape was for a message; the 
other held the stream of "key." 

(T0//0I) Fo rtunately for the British and the 
history of computers, that binary stream present- 
ed an opportunity to avoid the difficulty of align- 
ing the tapes. When it was realized that the sec- 
ond Robinson tape was a stream of algorithm- 
generated bits, it was suggested that a machine be 
constructed that substituted tube circuits for the 
additive tape. Reckoning that the number of 
tubes needed for the generation of binary combi- 
nations was reasonable, GC&CS gave the green 
light to construction of the Colossus. 83 

(U) The Colossus was something of a miracle 
of project management. It took less than a year to 
create what many consider the finest electronic 
pre-computer. Colossus kept 2,500 tubes and a 
high-speed photoelectric paper tape reader in 
synchronization. It could even be coaxed into per- 
forming some primitive program steps and "if 


Page 97 


(TO//OI) Colossus was very smart. It fol- 
lowed some ofTuring's maxims about how 
to save search time. It had an electronic cir- 
cuit that polled its counters to see if the 
results of the run were building to a signifi- 
cant outcome. This "sigmage" circuit saved 
hundreds of hours of machine and analyst 
effort. It also prevented Colossus from print- 
ing the result of every pass of the tapes, 
something the wasteful Comparators and 
the first Robinsons could not do. 84 

(U) The first of more than ten versions of 
Colossus was put into operation in January 
1944- 85 

(U) The Americans Almost Beat England 
to Electronics 

(U) The U-boat rampage in the Atlantic 
led to extreme criticisms of the American 
navy while the army was becoming worried 
that its men would go into battle in Africa 
and Europe without an"E" capability. 
Politically dangerous, Britain was giving 
them far less than the already meager ration 
of Enigma information it was providing the navy 

(U) Responding to all the various pressures, 
OP-20-G put even more resources into its frantic 
effort to conquer the U-boat "E," and the army 
began to think of the machines it might need for 
what it called the "Yellow Problem." 86 The army's 
SIS had difficulties obtaining information and 
resources and did not launch a machine program 
until the fall, but at the beginning of summer 
1942, OP-20-G hinted it had a solution to the M4. 
Within another two months it announced that its 
men had beaten Britain and the great Wynn- 
Williams to the creation of the heart of a fully 
electronic Bombe. 87 The circuit wasn't the univer- 
sal machine "G" wanted, but no time was lost in 
trying to exploit the development. NCR was taken 
over by the navy to be a research center and pos- 
sibly a production site. John Howard's old group 
became an integral part of Howard Engstrom's 

(U) Colossus 

"M" as it was reorganized to oversee the electron- 
ic Bombe work at NCR. 

(U) Wenger, Engstrom, and their like had to 
show some results. 88 The Bombe became impor- 
tant to "M's" survival as well as to the Battle of the 

(U) No Time for Electronic 

(U) In late summer 1942 the engineers of the 
"M" group decided their work was far enough 
along to submit it to an experienced production 
engineer for examination. Of course, they turned 
to Joe Desch. He spent almost two months exam- 
ining their bench model and their designs for an 
electronic Bombe. 89 He came to a devastating 
conclusion: An electronic Bombe was an impossi- 
bility! 90 A universal machine would need thou- 
sands more tubes and even higher speeds. The 

Page 98 



thousands of tubes would be difficult to acquire, 
would create too much heat, and would demand 
more electrical power than could be supplied. 91 

(U) A Crisis of Organization and Technology 

(U) Desch commanded so much respect that 
the responsibility for a new design was shifted to 
him. Necessarily, he was informed of one of 
America and Britain's great secrets, Ultra. After 
additional study of what was known about the 
Turing Bombe, he promised that he would be able 
to produce an electromechanical machine that 
could tackle the Shark M4. He declared he could 
create an original American Bombe but a non- 
electronic one. 

(U) Immediately, anew effort, the second 
American Bombe project, was begun. As a result, 
Wenger's dream of a Rapid Machine program 
was saved. For most of the remaining war years, 
the electromechanical Bombes devoured the 
energies of "G's" engineers. To fulfill the commit- 
ment to Desch's necessary backward technologi- 
cal leap, all the truly advanced projects and ideas 
were made stepchildren. 

(U) The second American Bombe project 
almost faltered, but it eventually became a tri- 
umph for OP-20-G and the American intelligence 
community. The success of the Bombes and the 
Allied work on machines for the Pacific war final- 
ly established the credibility Wenger needed to 
try to make research a permanent part of OP-20- 
G's peacetime operations. 

(U) Searching for a Place in Ultra 

(U) The second American Bombe project was 
part of an attempt to readjust the relationship 
between Britain and America's codebreakers. 
Desch may not have known it, but his Bombe was 
essential to OP-20-G gaining a greater role in the 
Ultra Secret and to becoming a producer of oper- 
ational information. Without an American 
Bombe, the United States would have remained a 

consumer of British-controlled Ultra informa- 
tion, and OP-20-G would have continued under 
the old understanding: Both nations could pursue 
independent (unaided) research, but Britain 
would control all operational activities. Although 
Britain had begun to ask for American help on the 
Atlantic problems, with the failure of the second 
Bombe project it would have been very reluctant 
to make OP-20-G an equal partner. 92 

(TS//0I) The friction with Britain over Ultra 
intensified soon after the approval of Desch's 
sketch of a modified British machine. The navy's 
men became intolerant of what they considered 
broken promises by the British about their four- 
wheel Bombe. 93 OP-20-G more than hinted that 
it would build as many as 350 of the Desch 
machines before spring 1943 when the U-boats 
were expected to launch a mass attack. America 
was notifying Britain that no matter what it took, 
the United States would win the U-boat war. 94 
The Americans declared they intended to build 
enough machines to test all Shark wheel orders 
simultaneously. 95 

(V) The Power of Innocence 

(U) Given all that the Americans did not know 
about the Bombes and all that was required to 
make them useful, miracles were required. 
Joseph Desch's first description of the proposed 
"G" Bombe and its powers reveals how much the 
American Bombe program was based upon the 
type of optimism that comes from innocence, if 
not ignorance. 


- ffS//0I) Some ofthe detailed plans ofthe 
three-wheel British Bombes had begun to reach 
the United States in late summer 1942, but 
Desch's design was his own. He had begun his 
plans before the British had revealed more than 
the bare essentials of their machine and the crib- 
based "menus" that made it work. ^ And he 
arrived at his first design before he had been able 
to test his assumptions about the way the essen- 
tial components of his Bombe would behave. 


Page 99 


£S //SI) Desch's plans were technically opti- 
mistic. He thought it possible to create a drive 
system that could keep twenty-four double-ended 
Enigma analogs (ninety-six commutators) in per- 
fect synchronization, A large electric motor would 
drive a high-speed shaft that would directly turn 
the shafts for the fast wheels. Gears, machined for 
complex ratios, would connect the high-strength 
rods for the slower wheels to the main shaft. The 
gears and shafts had to be of the highest quality 
material and workmanship to stand the stress 
placed on them when the machine suddenly 
stopped and restarted. 

fS^That was one unique feature of the pro- 
posed American Bombe that would put Desch's 
faith in mechanics to a severe test. He proposed 
an automatic stop, rewind, and restart system. 
That would save critical running time and, as 
important, avoid having the machine's operators 
having to hand-crank the device when a "stop" 
was sensed. 

£S / - /SI) Once the testing circuits identified a 
possible "hit," his machine would cut the power to 
the main shaft, apply brakes and bring the com- 
mutators to a halt all within a fraction of a second. 
Then a second motor would immediately drive 
the commutators backwards until another circuit 
signaled that the possible "hit" position had been 
reached. The machine would then perform anoth- 
er series of circuit tests, including a diagonal 
board search. If those tests indicated there were 
no contradictions, the commutator positions and 
the diagonal board indicators (the "story") were 
printed. Fortunately, "G" had not asked Desch to 
build a machine that tested for the ring settings as 
well as for the wheel orders and steckers. 

— ( 8 //3I?1 mmediately after the "stories" were 
printed, the motor would be restarted, the clutch 
on the central high-speed shaft would be 
engaged, the gears would mesh, and the commu- 
tators would turn in synchronization until anoth- 
er set of wheel positions indicated an Enigma set- 
ting that might have produced the crib. 

-*&H)esch did not seem to worry about the 
stresses that the quick stop, rewind, and restart 
systems would put on the shafts and gears, but 
Alan Turing certainly did. When he visited NCR 
in December, he warned "G" that it would be 
unlikely that any machine could be kept in work- 
ing order when it was asked to defy the laws of 

(6) D esch did not back away from the auto- 
matic rewind system nor did he change his mind, 
during fall 1942, about having three complete 
Bombes in a single rack. Such a configuration 
would save precious space (336 Bombes with 
32,000 commutators in 112 racks). He was con- 
vinced that the frames would tolerate the heavy 
vibrations from the three machines, which would 
be independently starting and stopping. 

-£ S) -Desch also kept his faith in the ability of 
American technology to make the "G" Bombe 
more flexible and many orders faster than those 
in operation in England. His Bombes were to be 
very rapid, several times the speed of the British 
three-wheel bombes and twice the speed of 
England's proposed four-wheel machines. 98 The 
fast wheel was to revolve at 3,400 rpm. The oth- 
ers would run at proportionately lower speeds, 
turning only when their faster mate had complet- 
ed a full turn. The second wheel, for example, 
would take one step after the fast wheel had made 
a complete twenty-six-point revolution, plus 
additional revolutions to compensate for the time 
the other wheels needed to turn over. 

-(S^-When desired, the Bombe could be turned 
into a three-wheel enigma analog, and, Desch 
hoped, it could be run at various speeds." 

(S^-Joseph Desch premised his design on the 
power of America's mass production methods to 
make all parts ofthe Bombes interchangeable. 
NCR's machinists had assured Desch that he 
could achieve his critical goal of having every 
commutator fit on any ofthe thousands of spin- 
dles on the Bombes. That was a critical feaUire. If 

Page 100 



commutators had to be tailored to each machine, 
his system would be impractical. That would 
demand too many highly skilled workers for com- 
mutator construction and too many commuta- 
tors, perhaps as many as 60,000. 

-f9}- Desch promised to make electronics as 
well as mechanics go beyond normal expecta- 
tions. He said he could overcome the problems 
that had led most engineers to avoid the use of 
tubes. Amultitube Rossi detector circuit would 
monitor the machines for possible "hits," another 
circuit would remember where the "hit" occurred, 
one would control printing, and, he hinted, 
another very complex one would handle the 
"diagonal" test. He was not sure in September 
1942 how many tubes each Bombe would have, 
but his memoranda hinted that his electro- 
mechanical machine would need perhaps as 
many as 1,500 electronic components. 100 The 
high-speed diagonal board alone might need 
more than 1,000 tubes. 101 

.^•His September design asked more than 
could be expected of gas-filled or vacuum tubes in 
the early 1940s. His faith translated into the blind 
hope that the navy's engineers could keep more 
than 300,000 tubes running atone time. "G" 
wanted to run each wheel order simultaneously. 
That meant 336 Bombes with perhaps 1,000 
tubes each running without a flaw for perhaps as 
long as an hour. 

■48^ The electronics posed a serious challenge 
to the navy engineers. They would have to find 
ways to handle the heat generated by the elec- 
tronic components and create methods of identi- 
fying troublesome tubes before they failed. Desch 
went ahead, believing that the "G" would find a 
way to overcome all the problems that had kept 
men like Vannevar Bush from trying to build 
large-scale electronic machines. 

4S> Joe Desch had to believe in the future of 
electronics; his machine could not work at what 
he thought was a minimum speed without elec- 

tronic switches. He could not build a "memory" 
for the machine out of high-speed commutators, 
relays were too slow for the "diagonal" test, and 
simple capacitors seemed unable to do the job. 

(U) The Power of Ignorance 

•ijSf Desch thought the navy needed so many 
Bombes that ran so fast because the Americans 
had not yet learned of the methods GC&CS used 
to select the Enigma wheel orders that had to be 
tested on the Bombes. GC&CS had discovered 
many ways of telling which wheel orders the 
Germans would not use during a crypto period. 
They had also developed many cryptanalytic 
techniques, which eliminated particular wheels 
and wheel positions. 

~f Q) —. As significant for the history of the 
American Bombe project, in autumn 1942 "G's" 
experts did not fully understand the methods 
Britain had developed to allow the Bombes to 
quickly beat "chance." As a result, they had con- 
cluded that the United States had no alternative 
but to invest millions of dollars in machines that 
were very inefficient. 

(S// 8 I) -The Bombes would be valuable only if 
used properly. When they were given enough 
information, they speedily reduced the number of 
Enigma wheel orders that had to be examined by 
the analysts. But if used improperly, they could 
not sort out the wheel orders and stecker settings 
that could have produced the cribs by accident 
from those settings that were "causal." With only 
relatively short crib-plain combinations to test, 
twenty or so letters, a Bombe with a weak menu 
might filter out only a small proportion of the 
incorrect settings. Desch, for example, feared that 
typical menus would force analysts to comb 
through a third of a million possibilities after a 
run of the 336 Bombes to locate the setting that 
was the true Enigma key. loz 

(U) Using analysts to search for Enigma set- 
tings was time-consuming and expensive. Tests 


Page 101 


took from seconds to hours, and some of them 
demanded skilled if not just very devoted person- 
nel. From the list of combinations that were not 
eliminated by the Bombes, analysts used other 
machines and hand methods to see which was the 
unique one that produced plain text from crib. 
Thousands of analysts would be needed if the 
Bombes did not eliminate all but a very few of the 
Enigma settings. 

-(S>To be useful the Bombes had to eliminate 
more than just those combinations that could not 
possibly have produced the crib-plain combina- 
tion; they had to filter those that were unlikely to 
have done so. The only valuable pay-off from the 
use of the Bombes was a very short list of very 
likely "keys." Unless the list was short, there 
would be no significant savings in time and man- 

£¥Sf The only way to that short list was 
through the location of long and accurate cribs 
and the creation of powerful "menus." Starting 
with relatively long cribs, a menu was built 
through searching for letter combinations and 
connections (closures) between plain and cipher 
that would allow the Bombe's circuits to differen- 
tiate "chance" relationships from those that were 
caused by the true Enigma setting. 

(T0//ai) By mid-1942 GC&CS's wizards had 
turned menuing into a mathematical art. They 
had discovered much since Turing's first insights. 
They had tables showing what types of menus 
were needed to produce the desired short lists; 
they found that wisely selected cribs reduced the 
need to run all thirty-two of the Enigmas in their 
Bombes; they could calculate how many time- 
consuming machine stop and circuit checks 
would be expected per menu; and they could pre- 
dict how many possible wheel order and stecker 
settings would be printed per run. 

(•^ES} Strong menus were a necessity. 
Otherwise, the Bombes might stop their wheels 
and demand a return to the possible "hit" position 

so frequently that runs would take several 
hours. 103 

(S#SI) The British had also learned how 
counterproductive it was to run the Bombes with- 
out having 100 percent accurate cribs (a goal that 
called for avast infrastructure) that yielded 
robust menus. They reserved their Bombes for 
menus that would produce no more than a hand- 
ful of "stories" per run. By mid-1942 they had 
learned how to select fifteen letter cribs that pre- 
vented the Bombes from stopping more than a 
dozen times and printing more than five possibly 
true "settings" during a run. When the Germans 
made the mistake of providing excellent cribs, the 
Bombe could identify "the" and only "the" set- 
ting. 104 

( S//SI> But Desch and the Americans had not 
learned enough about cribs and menus by 
September to envision or wait for Bombes that 
would point to only a handful of possible solu- 
tions. In September Joseph Desch estimated that 
his ultra-high-speed Bombes would take, using 
typical menus, thirteen and one-half hours to test 
one wheel order on one Bombe. The reason: He 
calculated that a Bombe would have to stop 3,000 
times per run, perform its circuit tests, and then 
decide whether or not to print its declaration that 
it had found a highly probable solution. On the 
average, Desch estimated, the Bombe would sug- 
gest that one of three stops had found awheel 
order that should be tested by an analyst. 105 

(S} If, as envisioned in early fall 1942, "G" was 
to run its 336 machines simultaneously and con- 
tinuously, the Bombes would spew out some 
300,000 "probable" settings twice a day. How 
"G" was to wade through all that was not 

(S3 Even with the use of the best possible 
menus, Desch's Bombes seemed to demand great 
manpower investments for the production of 
timely intelligence. The navy's commitment to 
becoming a partner in Ultra is underscored by the 

Page 102 



acceptance of the consequences of Desch's most 
optimistic menu scenario. After consulting with 
OP-20-G's Enigma expert, Lieutenant Ely, Desch 
held out some hope that "G" would eventually be 
able to provide menus that would cut each 
machine's running time to three hours and the 
number of its stop-rewind sequences to less than 
700 per run, per machine. That meant that "G's" 
experts would need to test some 40,000 prints 
each day. 106 

^-The Americans were desperate. They 
accepted the Bombe program despite Desch's 
estimates. They were willing to invest millions in 
hardware and more in manpower for a system 
that was very inefficient. Despite Desch's esti- 
mates of how long the Bombes would take to pro- 
duce so few results, his report on the design of the 
American Bombe was quickly approved. 107 

(U) The Cousins Will Have Their A 

(U) Although they questioned the ability of 
"G" and its engineers to build a significant num- 
ber of Bombes and to devise the menus needed 
for them, the British had no choice but to take "G" 
seriously and to make the best of the situation. 
They quickly dispatched another cryptologic del- 
egation to the United States. 

(0//0I) Accepting what seemed to be the 
inevitable, GC&CS agreed to help the Americans. 
But they continued to argue that European intel- 
ligence should be left to them. The British 
explained more of their methods of avoiding the 
need to run all the Bombe's wheel combinations 
and orders to test a message. They dropped hints 
that "G" could expect a steady flow of valuable 
cribs and solutions. 

4 S//B¥ ) The British seemed to be even more 
generous when they agreed to a new arrangement 
in the Pacific. But they gave up very little and 
gained much by allowing OP-20-G to run the 
cryptanalytic and intercept operations in the area. 

(S//Sr) In exchange for Britain closing down 
some minor centers and allowing the Americans 
to direct the codebreakers in Australia, the 
American Navy promised to send, if practical, all 
raw and processed information about the Pacific 
to the British. Given how few resources Britain 
had in the region, it gained more than it relin- 

t&ffST) The Americans offered even more. 
They volunteered to provide GC&CS with copies 
of "G's" newest RAM devices and to train its tech- 
nicians in their use. 108 

( S // &I) In contrast, the Americans obtained 
much less than they hoped for in the Atlantic. It 
was agreed that the British would accede "to U.S. 
desires with regard to work on the German sub- 
marine and naval problem," but Britain, in effect, 
would be "the coordinating head in the Atlantic 
theater as the U.S. will be in the Pacific." 

(S//S+) The previous pledge to give "G" advice 
on analytical machinery was reaffirmed, and it 
was agreed "in principle" that Britain would col- 
laborate with "G" and send needed cribs, menus, 
and intercepts as long as the security of Ultra was 
not endangered. 109 

( S//S * ) In response to Britain's bowing-out in 
the Pacific, to its providing more knowledge of 
Bombe techniques, and to its obvious determina- 
tion to keep control of the European Ultra, the 
Americans tentatively agreed to build only one 
hundred Bombes." 

(S3 One hundred in this case meant a total of 
100 four-wheel Enigma analogs, in contrast to the 
Desch plan to have 100 racks with three Bombes 

(TS//6I) They also agreed to keep their 
Bombe design very fluid so they could respond to 
emergencies. "G's" technicians were also made 
aware they might be asked to play a backup role 
for GC&CS. Like the rest of Britain's military, 


Page 103 


GC&CS began to think it could depend upon 
America's industrial capabilities. In late 1942 
there were indications it might be forced to. "Doc" 
Keen's factories were stretched beyond their lim- 
its. America would be needed to handle cryptana- 
lytic machine emergencies. 1 " 

(U) "G" seemed ready to accept that role. It 
also did not reject Britain's suggestion that all 
Ultra-based naval actions be coordinated ones. 
There certainly was no hesitation when GC&CS 
asked the navy for a firm pledge to do everything 
needed to make Ultra America's most guarded 
secret. 11 " 

(U) The October 1942 negotiations did lead to 
Britain giving "M" a somewhat greater opera- 
tional role, or at least preparing them for one. 
GC&CS's representatives set "M's" men to using 
hand methods on various German systems and 
gave them more instructions on how to prepare 
menus (setups) for the Bombes." 3 During the 
next few months more and more technical details 
about Ultra flowed to the United States, and more 
ofEngstrom's bright young men in"M" traveled 
to England to work at GC&CS. But the British 
retained the power to decide what information 
would and would not leave Bletchley Park. 114 

(U) A Long Apprenticeship 

(TS// S I) A year after the October 1942 agree- 
ment, at the end of 1943, "G" remained an 
appendage to Britain's European Ultra. In 
November Howard Engstrom traveled to 
England to hear something quite like a lecture 
about "M" wasting valuable resources by running 
the Bombes on very weak American-devised 
menus. He had to agree to a return to using 
British cribs and menus. It was not until much 
later in 1944, after "M" had enough good inter- 
cepts, and after its cryptanalysts had honed their 
menuing skills, that "M" was granted effective 
independence concerning the Atlantic U-boat 
problem." 5 

(U) It took a long time for "G's* men to gain 
the necessary skills and to build an effective anti- 
Enigma organization. There was progress, but it 
came slowly. In early 1943 American cryptana- 
lysts were applying GC&CS paper and tabulating 
machine techniques to crack some German mes- 
sages intercepted by the British. 

(TS//SI) But most of the first half of 1943 was 
a long practice session for the Americans. Using 
GC&CS-supplied keys, they deciphered and ana- 
lyzed the large American backlog of intercepted 
Atlantic traffic. Then GC&CS forwarded new mes- 
sages and their keys. 

■ £T0//3fi To aid the deciphering process, the 
navy's men in Washington built anew electro- 
mechanical device, the M8. The M8 was not an 
analytical machine, but soon after its appearance 
in October 1942 it became an invaluable tool for 
"G's" analysts. More refined models began to 
appear in spring 1943. The M8s were reworked 
versions of the Americans' own automatic wheel- 
based encryption machine, the ECM. The navy 
yard's engineers added a plugboard, Enigma 
wheels and Letterwriter equipment to turn the 
ECM into an automatic and relatively high-speed 
"translation" machine. Once the wheels and plug- 
board were set, the "stories" from the Bombes 
could be rapidly tested, or entire messages could 
be deciphered at rates up to 600 letters per 
minute. A few months later the M9, another sim- 
ple Enigma analog, appeared. The M9 was a very 
sparse combination of wheels and plugboard that 
was extremely useful once the Bombes were in 
operation. An M9 was later placed near every set 
of Bombes, allowing their operators to make 
immediate checks of printouts and to locate miss- 
ing Enigma plugboard settings." 6 

CTS//6D But much of the American Ultra 
effort of 1942 and 1943 proceeded without the 
help of automation. Enormous human resources 
were put into a paper version of a bombe: a 
1,000,000-page catalog that could be used to 

Page 104 



drag a short crib through wheel settings to find 
possible Enigma keys." 7 

(U) In late 1942 Alan Turing began a series of 
visits to America telling "M" more about Ultra 
and the construction efforts on the latest British 
Bombe." 8 The two nations' intelligence services 
and the two Bombe projects became closer as 
1942 ended. But Desch's Bombe was an American 
product." 9 

■6S) By the time he received adequate informa- 
tion about Wynn-William's work and that of 
"Doc" Keen on his four-wheel "Mammoth" 
Bombe, Desch had his design relatively fixed. 120 
He was convinced that the automatic rewind fea- 
ture was essential, and he did not wish to halt his 
work while the British proved that all "hit" circuit 
tests could be done without rewinding the wheels. 
He was convinced that his steel shafts would be 
more reliable than belt drives and that menu set- 
ting should be done with a set of switches rather 
than clusters of hand-inserted Jones plugs. 

(U) Desch and his NCR men and the young 
engineers in John Howard's "M" group took the 
very heavy responsibility of creating a unique 
machine and a path-breaking production line to 
defeat the U-boats. 121 

(U) Desch Takes Charge 

(U) As soon as his general design was 
approved in September 1942, Desch began to 
refine his ideas and looked forward to immedi- 
ately building a prototype. 122 At the same time, 
the old engineering group from MIT was ordered 
to put the other OP-20-G machine projects on 
hold until the critical Bombes were ready. 
Financing was not a problem, and the second 
American Bombe project, which atone point 
would employ over 1,000 manufacturing work- 
ers, received the highest priorities for personnel 
and material. The Bombe project had its own 
building in Dayton with armed Marines and spe- 

cial secret rooms to manufacture 

and use the 

(U) Wenger Gets His Organization 

(U) The Bombe was so important that the 
Bureau of Ships had to grant all the wishes of 
Wenger, Engstrom, and Desch. OP-20-G was 
able to convince the Bureau to create a new 
administrative organization for all the high-speed 
machine projects: the Naval Computing Machine 
Laboratory (NCML) at Dayton, Ohio. Informal 
terms the Bureau's NCML was the boss of the 
Dayton work, but by early 1943 it was really a 
support organization for OP-20-G's group of 
engineers and scientists. 

(U) The "M" group was also gaining power. 
The country's best mathematicians, physicists, 
and engineers were brought into OP-20-GM. 
That allowed Engstrom to have a self-contained 
machine development group that easily chal- 
lenged the Bureau's technical authority. Of 
importance to the nature of the postwar RAM 
program, the "M" engineers were integrated with 
the NCR workforce. That gave the machine 
designers the freedom to merge research and pro- 
duction and, combined with the virtual takeover 
of NCR, it allowed Wenger a constant interaction 
with and power over the manufacturing process. 

(U) Of Tires and Transmissions and a 
Disappearing Laboratory 

(U) Such freedom and the massive resources 
the navy was willing to pour into Desch's project 
were not enough to sustain the hopes of 
September and October 1942, however. At the 
opening of 1943 a prototype of his Bombe had not 
been assembled, and there were serious questions 
about the practicality of the components that had 
been constructed. 123 The rejection of the plan for 
an electronic machine and the reversion to the 
electromechanical technology of the British 
Bombes had not led to the easy solutions "G" had 


Page 105 


expected. The Bombe and Rapid Machine proj- 
ects were in trouble again. 

(0//SI) Joseph Desch's first designs had 
called for a Bombe that was a close analog of the 
Enigma, but it was to be a very, very fast one. 
There were to be more than twenty sets of four 
wheels each in a Bombe. Each of the four wheels 
was to be of the same size and was to be con- 
structed out of typical materials of the era. The 
Bombe's commutators were to be made of either 
hard rubber or Bakelite, standard insulated hous- 
ings of the 1940s. 

(U) Inserted within the inner face of those 
wheels were rather large copper contact bars. Joe 
Desch knew he would have to make them of spe- 
cial lengths and shapes to prevent spurious elec- 
trical contacts from being registered as "hits." 

(S//SI) The fastest of the four wheels was to 
spin around more than sixty times a second. That 
rate of speed seemed essential. And also essential 
was the complex gearing that would be required 
to pace the movement of the slower commuta- 
tors. The gearing requirements included the diffi- 
cult-to-machine-and-maintain Geneva gears and 
a stepping control system that reminded one 
experienced engineer of the complexity of the 
recently invented automatic transmissions for 
automobiles. 124 

(G//0I) Although a challenge, the group at 
NCR had few doubts about creating the Bombe. 
September's optimism about such mechanical 
and electrical parts did not last long, however. 
The first serious disappointment came quite 
soon. It was found that the commutators could 
not tolerate such high speeds. The fast wheels 
were blowing apart. The problem could not be 
overcome, Desch concluded, so he significantly 
altered the design of the Bombe. 

(S//SI) By December he envisioned a 
machine that would have two small "fast" wheels. 
The smaller wheels, he hoped, would rotate at 

least at 1,800 rpm without disintegrating. Soon 
he had to admit to other problems. He warned 
Engstrom that the commutators might not be 
interchangeable from Bombe to Bombe. 
Production difficulties might cause something 
worse. The commutators, Desch said, might have 
to be permanently attached on each spindle. 125 

( T S //SI) The decision to create a Bombe with 
two fast wheels created near panic in 
Washington. Since any of the Enigma wheels 
might be assigned the "fast" position, each of the 
eight known Enigma commutators would have to 
be cloned by two, not one Bombe commutator. 
Howard Engstrom let Desch know in the 
strongest terms that he disapproved of the two- 
wheel design. It would create a logistics night- 
mare, wrote Engstrom. More than 40,000 or 
50,000 ofthe expensive commutators would 
have to be immediately stockpiled and made 
available for use. If the Germans altered the 
wiring on their wheels or added new ones, no 
manufacturer could respond quickly enough to 
produce the new wheels. 126 

(U) Desch promised that he would do his best 
to make the wheels interchangeable from Bombe 
to Bombe, but he could not guarantee that he 
could produce a Bombe of any significant speed 
without the dual fast wheel feature. 

CT0//0E) Then a very great gamble was made. 
Washington declared that a solution be found. 
The American Bombe would have only one size 
wheel! Desch and his men had their orders, but 
no solutions. They began an intensive search. 
Their reward was disappointment. Prototype 
after prototype kept disintegrating when put at 
the high-speed position even when the revolu- 
tions per minute were reduced to less than half 
than originally planned. 

■£ T$//SI) -The answer eluded everyone. It was 
not until some ofthe young officers stationed at 
NCR realized the similarities between the com- 
mutator's problems and those of automobile tires 

Page 106 



that there was a glimpse of hope. How were tires 
able to hold together during auto races? The 
answer seemed to lie in a new product, rayon. The 
officers learned that webs constructed from it 
were being used to reinforce new types of rubber 

(TQ//3I) There were visits to local tire compa- 
nies and some tests. The situation appeared 
hopeful, but no one was sure that the experiments 
could be translated into a mass production sys- 
tem for the commutators. 127 

£ TS//GF )- Because they had no choice, Desch's 
team went ahead with the rest of the Bombe proj- 
ect while they waited for word about the commu- 
tators. As they did, they confronted another prob- 
lem whose solution was also tied to the automo- 
bile industry. 

fflS//Sfl The gearing system for the Bombe 
proved more complex and temperamental than 
expected. No one seemed able to correctly align 
the components. Finally, the more senior engi- 
neers asked some of the younger men in Dayton if 
they had any experiences with gearboxes. One, 
whose engineer father had worked for Tom 
Edison, had some hands-on experience with the 
new automatic transmissions. When he exam- 
ined the Bombes, he saw much that related to his 
previous experience. He volunteered to try to 
solve the gear assembly problems. 128 

( TS// S I) There was another important prob- 
lem with the Bombes that almost halted develop- 
ment. Desch had rejected the idea for a fully elec- 
tronic bombe, but had no alternative other than 
to rely upon electronics for many parts of his 1942 

(TS//SI) Desch's September design suggested 
a need for perhaps as many as 1,500 tubes in rel- 
atively complex circuits. The fast diagonal board 
to test for stecker settings might call for over 

1,000 tubes. Given the size of tubes of the early 
1940s and the heat they generated, an alternative 
to off-the-shelf technology had to be found. 
Desch's past experience led him to believe that 
small multipurpose tubes might be created. He 
made some attempts to refine his previous 
designs, but his many other responsibilities 
pulled him away from the needed solution. He 
had to have help. Fortunately, the NCR project 
had such high priority and such vast resources 
that one of the nation's leading tube experts could 
be summoned to Ohio and allowed to order 
everything he needed to create an advanced labo- 
ratory. His work proved successful, and he was 
able to deliver the specifications to manufacturers 
for the special tube. It was a tiny four-in-one tube 
that became the basis for the ultra-fast diagonal 
board. It would be produced in carload lots, and it 
reduced the number of separate tubes in the 
Bombes to fewer than 500. 129 

(U) While the "G" group waited for the solu- 
tions to the commutator, gearing, and tube prob- 
lems, they faced a very chilling possibility. For a 
time, some in "M" worried that their opportunity 
had been lost because GC&CS was able to reenter 
the Atlantic U-boat system at the beginning of 
1943 without the use of any four-wheel Bombes. 
The reentry came through the capture of docu- 
ments from a U-boat and the discovery of some 
very sloppy procedures on the Shark network. As 
a consequence, the British were able to read the 
four-wheel Enigma messages using their old 
Bombes and hand techniques. 

(U) But the British and the Americans soon 
realized how temporary the new solution was. As 
the spring U-boat offensive opened, the Germans 
changed some of their codes and tightened up 
their procedures so that the Allies were again shut 
out ofthe submarine systems. They remained 
blind for a frightening ten days during what 
became the worst month in the history ofthe bat- 
tle ofthe Atlantic 


Page 107 


(U) Samng the American Bombe 

(U) At least three months before that ghastly 
March 1943 U-boat slaughter, OP-20-G realized 
that Desch's machine was in serious trouble. 
Pressure was put on the staff at NCR to work 
overtime. Joe Desch was told to drop his many 
other electronics projects for the NDRC, 
Aberdeen, and the army. And the navy went over 
the head of the new president of NCR and wrote 
directly to Colonel Deeds to make sure that NCR 
gave the Bombe project all it needed. Under prod- 
ding from the Chief of Naval Operations, Deeds 
quickly ordered Dayton to devote less time to its 
other and more profitable war work and give the 
Bombe all of its attention. 130 

(T8//SI3 The first design for the pilot model of 
the Bombe was submitted in January. Joe Desch 
and John Howard responded to British sugges- 
tions and incorporated them in a second design 
even as they rushed to construct the first proto- 
type. But the men in Dayton were not keeping 
pace with the war. 131 As the great Atlantic battle 
began in March, all that had emerged from some 
seven months of work were two wheezy prototype 

CES//Sff Their commutator racks sat on saw- 
horses, and their other components were scat- 
tered around the workroom, connected by scores 
of wires that were soaked with the oil that flowed 
out of their drive shafts' housings. Their commu- 
tators continued their obstinacy, and the crew of 
engineers endured repetitions of lowering the fast 
wheels' "rpm," then having to dodge their frag- 
ments as they splintered. No one was sure that 
the two models, Adam and Eve, would prove 
themselves and serve as test beds for the vital 
production machines. Nonetheless, Washington 
decided it could wait no longer and in early April 
Desch committed to a final design for the produc- 
tion version of the American Bombe although he 
was not sure that it would work. 132 

(U) There was progress at Dayton during 
April and May, but no machines! The group at 
NCR could not even tempt the two Bombe proto- 
types to run for more than a few inadequate min- 
utes. Fortunately, escort carriers, airborne radar, 
a central command center for subhunting (the 
Tenth Fleet), and changes in the once vulnerable 
Allied convoy codes began to bring the Atiantic 
under control. Enigma cracking played its part, 
but not through the promised American techno- 
logical wonders, the Bombes. 

(U) A Bombe Too Lale 

(U) Order was restored in the Atiantic before 
the first American Bombe was even put to its 
tests. The problems in the Atlantic and the com- 
ing European offensives called for another read- 
justment in the rules for cooperation in the intel- 
ligence field. The BRUSA agreement made the 
United States Army a partner in the Ultra Secret, 
but a very junior one. OP-20-G and the United 
States Army again agreed to focus on the 
Japanese problems and to allow GC&CS to deter- 
mine what the Americans would do or would not 
do against the Enigma and Fish systems. 

(U) As the mid-May 1943 negotiations came 
to a close, Joseph Wenger remained unsure of 
OP-20-G's future. Even if "G" was too late to be 
the savior of the Atiantic, there was still much to 
do to counter the U-boats. The German Army, Air 
Force, rocket development team, and police agen- 
cies showed signs of changing over to four-wheel 
Enigmas. And, in the systems continuing to use 
the three-wheel machines, anticipated alterations 
in procedures and in the use of their plugboards 
threatened another round of crises. If the 
American Bombes could be made to work, they 
still might play a significant role. 

(U) In late May Wenger ordered Desch to 
allow the two temperamental prototypes to be 
used on messages sent from Washington. The 
results were to be forwarded to the British as 

Page 108 



examples of American abilities. 133 Howard 
Engstrom, in charge of the new Enigma message 
work, felt defeated when Adam and Eve refused 
to run for more than a few hours without spurting 
oil or developing incurable cases of faulty electri- 
cal contacts. 134 

(U) A Program Based on Another 
Technological Bet 

(U) Adam and Eve continued their tantrums 
as June approached, 135 and the production model 
was yet to be assembled. The tension mounted 
when it was learned that as Dayton again faltered, 
Britain completed its first four-wheel Bombe, put 
its first tape and electronic Robinson to use, and 
began the construction of the advanced electron- 
ic COLOSSUS. 136 

(U) Adam and Eve, the prototypes, were in 
too much trouble and were too vital to working 
out critical technical problems to be used by the 
cryptanalysts in Washington. Desch's crew and 
Howard's NCML engineers put in longer hours 
using Adam and Eve to unravel the problems with 
the parts for the production machines. The com- 
mutators were reworked and the drive-mecha- 
nisms altered. 137 By mid- June there were hopes 
that all the problems had been conquered. 

(U) However, the production crew had still 
not released the first two copies ofthe final 
model, Cain and Abel. Desch pushed his people 
harder, and the NCR factory began to assemble 
components at an even faster pace. They could 
only hope the parts would function when put 
together in the Bombe. 

(U) July 26: a Day of Defeat 

(U) It took Joe Desch another month to send 
the first two production models to the test 
floor. 138 Then he was able to have thirteen more 
ofthe new Bombes assembled by the last week of 
July, but none would work! 139 July 26, 1943, was 
a critical day in the history ofOP-20-G and the 

NCML. At the very last minute, Desch made a dis- 
covery that revived hope. Running the Bombe's 
Bakelite code wheels at extreme speeds was again 
causing invisible distortions leading to false elec- 
trical contacts. Desch predicted that careful stor- 
age, handling, and refurbishing would solve the 
problem. 140 Apparently, Desch had replaced the 
small fast wheel on the first prototypes with ones 
the same size as the others to please Engstrom. 
Again, his judgment was trusted. The wheels were 
reworked and production was resumed based on 
his hope that the last-minute modifications 
would provide a permanent cure. 

(U) A Victory, a Bit Too Late 

(U) Despite all the false starts, delays and 
problems, Desch built one ofthe most complex 
machines in the world. The 1943 Dayton Bombe 
was a seven-foot-high, eight-foot-long, two-foot- 
wide and 5,000-pound marvel. It housed sixteen 
four-wheel sets of Enigma analogs and the 
Welchman diagonal board. Its sixty-four double- 
Enigma wheel commutators each contained 104 
contact points, which had to be perfectly aligned 
when they touched the copper and silver sensing 
brushes. Such alignment and synchronization 
were difficult to achieve, especially for the fast 
wheel. The achievement was more remarkable 
because Desch was able to keep his promise of 
making the commutators interchangeable. 

(U) There had been some compromises in 
order to convince the machine to work. It was 
much slower than hoped for. Fewer than 2,000 
revolutions a minute had to be accepted because 
even the reinforced commutators could not stand 
up to higher speed. And running the Bombe's 
main shaft even at the lower "rpm" without creat- 
ing the sparks and short circuits that ruined a test 
continued to be a problem. 

(TS//SI) It was quite embarrassing to have 
had to install a conduit system under the machine 
to catch the oil that was sprayed on the main shaft 
to keep it from overheating. And the engineers 


Page 109 


did not like the idea of having to pour a quart or 
so of oil into the machines every day. 141 

( T S //8I ) There had been some other compro- 
mises. The NCR devices did not incorporate a 
means of producing irregular stepping of the 
slower wheels, and the summer 1943 Bombes 
were not the compact three complete units-per- 
frame devices Desch had sketched in the previous 
year. Trying to balance the engineering demands 
with cryptologic power had led to the Bombes 
being composed of sixteen, not twenty, units, and 
having only one Bombe per frame. 

(TS//0I ) Sixteen "Enigmas," one diagonal 
board, and fewer testing circuits made the 
American Bombes much less discriminating than 
the new British ones (thirty-six Enigmas and two 
boards); but Desch's Bombes were much more 
reliable and needed remarkably little mainte- 
nance once they were broken in. 1 * 2 

( T9//0I ) Although Desch's model was based 
on the logic, parallel architecture, and hardware 

of the British Bombe, his machine was an origi- 
nal. The truly distinctive part of Desch's machine 
was its electronics. 

( TS// S I ) He did not rely upon the designs of 
the new Keen Bombe, the Mammoth, nor did he 
copy Wynn-Williams' ideas. 143 

( TS//6I ) Although Wynn-Williams' partially 
electronic machine, the Cobra, was more sophis- 
ticated in some ways than Desch's, Desch's was 
more effective. Wynn-Williams' device printed 
solutions on the fly, but it proved somewhat unre- 
liable. Desch's Bombe also proved more trustwor- 
thy than Keen's latest electromechanical one. 144 

( T6//83 Desch was forced to bow to some 
other technological limits since his Bombes con- 
tained sixteen not twenty-four (or, like the British 
machines, over thirty) Enigma analogs. Each of 
the sixteen units was housed in a separate rack, 
which took up as much space as the three-per- 
rack configuration that had been the goal in 
September 1942. 

( TG//SI) Boifoe 

(TS//SI ) Although the 
American Bombes did their 
jobs, they could not use long 
and discriminating cribs of 
more than sixteen letters, as 
could the British devices. 
Desch had balanced his under- 
standing of the power of short 
cribs against the mechanical 
difficulties of driving a large 
number of commutators. U5 

( TS//SI) The information 
about cribs and menus that 
had been revealed after 
October 1942 also helped 
Desch decide against attempt- 
ing to incorporate a helpful but 
challenging feature; automatic 
slow- wheel turnover. Desch 
did not build a means of kick- 

Page 110 



ing the slower wheels to a new position relative to 
its mates after a faster wheel had completed a full 
revolution. Relying upon the probabilities that 
hits could be discovered before the fast wheel 
reached the point where the next wheel should be 
pushed ahead one or more letters, hand-setting of 
displacements during the commutator setup, and 
the use of a two-part menu allowed the Bombe 
design to remain manageable. 1 * 6 

(T8//BI) However, this meant that when 
"hoppity" menus (fast-wheel turnover positions 
known) were run, the Bombe operators had to 
stop the machines, then reset wheel positions by 
hand and restart the Bombe. 147 

(U) Ignorant No More 

(U) However, with the help of his electronic 
memory system and the maturation oPG's" 
Enigma cryptanalytic skills, Desch came very 
close to achieving all the hopes for the American 

(T S // S I) Because of his engineering skills and 
the use of strong menus, the NCR Bombe took 
twenty, not fifty, nor the worst case, 380 minutes, 
for a run. A major reason was that the British 
cryptanalysts (and later "G's" experts) were able 
to supply menus that produced on the average 
five- or so prints or "stories" per run -not 
40,000.*-* The menus eliminated so many of the 
possibilities that the Bombes stopped, rewound, 
and restarted very infrequently. That saved sig- 
nificant amounts of time. 

6RB/SI) In terms of raw speed, Desch 's 1943 
machine was 200 times faster than the Polish 
Bomba, at least twenty times faster than the 
Turing Bombe, and at least thirty percent faster 
than Britain's 1943 four-wheel Bombe. 149 His 
machine was able to run either three-or four- 
wheel tests. 

(TO//OI) As important for the success of the 
NCR Bombes were the menus for them. 

Unfortunately, it took the Americans many, many 
additional months before they learned how to 
consistently supply "strong" menus. Fortunately, 
by late summer 1943 the British were willing to 
wire cribs and other Bombe instructions directly 
to "G's" Washington headquarters. They forward- 
ed them as soon as they located the reencode- 
ments and other German procedural errors that 
were allowing them into the Shark system. 150 

(U) The Bombes at Work 

(U) Desch's manufacturing techniques gained 
the respect of the once skeptical British. By mid- 
November Washington had over fifty bombes in 
operation and thirty more on site, 151 The 
American navy finally began to be a truly produc- 
tive Ultra member. By the end of the year, the first 
contract was completed, lsa and Engstrom began 
to turn his crews to other technical and cryptana- 
lytic problems. 

(U) Although the second American Bombe 
project, from the first investigations to the last 
delivery, took almost a year longer than expected, 
Desch and OP-20-G received applause, not criti- 
cism, in late 1943. 153 Asa result of Dayton's 
achievement, the British found it impossible to 
continue on with a condescending attitude. The 
Americans soon became the guardians of the U- 
boat work, and Britain felt confident enough to 
concentrate on the Fish system and German army 
traffic. 154 From mid-1943 to the end of the war, 
M4 was open to America and Britain. 155 

(U) But the American Bombes were born a bit 
too late. By the time the Washington center 
received its machines, the four-wheel U-boat traf- 
fic was light. 

( TS//8f j Or at least the British thought so. 
They complained that the Americans were run- 
ning their precious Bombes against low-priority 

messages and were using menus that were highly 
unlikely to produce a break. They more than sug- 
gested that Engstrom should agree to a true 


Pagt 111 


"sharing" of Bombe resources. They wanted the 
American navy to run important German air 
force and army problems and to help with the 
"European" part of the war. 156 

(Tfl//(g> Productive work was found for the 
more than one hundred American machines. The 
navy's men soon began the analysis of other 
German Enigma systems. Although somewhat 
worried about breaching the agreement with 
Friedman's American army group, OP-20-G took 
on much German three-wheel air force work. 157 

Q£8d By autumn 1944, 60 percent of the navy's 
Bombe time was devoted to German air force and 
army problems as presented by the British. 158 
OP-20-G contributed more to the non-naval 
work. After rejecting a British plea to quickly con- 
struct another fifty Bombes and complaining that 
England had not contributed enough to the four- 
wheel effort, NCR built another two dozen 
Bombes. The new machines were somewhat more 
sophisticated than the first version. 159 Britain did 
place more and more responsibility for four- 
wheel bombes on the Americans, even announc- 
ing at one point that Keen would return to build- 
ing three-wheel machines. The second series of 
American Bombes had a "double input" feature 
that eliminated more false hits. 

(U) In addition to constructing the new 
Bombes, NCR built a series of attachments for 
them and the older machines. As well, they put 
into use other devices needed to automate the 
final steps in identifying Enigma keys. 

( T S // S D An Enigma attack was aimed at dis- 
covering seven things about an Enigma setup: 

1. Rotor wiring 

2. Reflector plugging 

3. Stecker setting 

4. Rotor order 

5. Window setting (starting point) 

6. Ring or core setting 

7. Notch pattern (turnovers) 


( TS//SI J- Given a strong crib of about sixteen 
letters, knowledge of the rotor wiring, a correct 
guess as to the reflector's plugging and a large 
number of Bombes, analysts could expect to get 
back some very good indications of the stecker 
setting, the rotor order, and some indirect help on 
the window setting, the ring setting, and the 
notch pattern. The Bombe was a relatively strong 
and quick means for a solution. But it needed 
some help. 

(U) More to It Than the Bombe 

(U) The Bombe was a powerful anti-Enigma 
tool, but by itself could not yield all that needed to 
be known about Enigma settings. On the other 
hand, it was, in some contexts, an example of 
overkill. When several elements of an Enigma 
setting were already known, other machines were 
much more efficient. 


(3S-) For example, Hypo, the microfilm 
machine that had been constructed so that a pure 
statistical attack could be mounted against the 
Enigma, was called into service as a "locator." 

( 3S//SF) When the Bombes had done their 
jobs and the rotor order and wiring, the stecker, 
and the Uncle Walter were known, Hypo was 
used to find the window setting through a letter 
frequency test. 162 Hypo helped in certain tough 
cases, but it took a great deal of time. Developing 
its film took forty minutes, and a "three-wheel" 
Hypo test took seventy minutes. And Hypo 
demanded humidity- and light-controlled 
rooms. 163 

(C//8T) The navy group at NCR found a faster 
way to handle such tasks. They built electrical 
attachments for the Bombes. These "Grenades" 
were large panels containing pluggable switch 
banks, which were used to control the Bombes. 
Their most frequent applications called for only a 
few of the sixteen banks on a Bombe and ran very 

Page 112 



quickly. In most instances a Grenade run took 
fifty seconds. 164 

(T0//0I) There were many varieties of 
Grenades. One of the firsts was for the window 
settings when the other Enigma crypto-variables 
were known. Mrs. Driscoll had outlined the logic 
for the first American Grenade to John Howard a 
month before the Bombe project had been 
approved in September 1942. He implemented 
those ideas in the technology of the Bombes as 
well as in the film machines. 165 

CES//S1) The Driscoll-Howard Standard 
Grenades appeared as soon as the Bombes came 
into operation. They were a great help because 
they reduced the effort needed to identify the 
window settings for succeeding messages once 
the Bombes had found the daily key. With the 
known daily key, all that was needed was a short 
four- or five-letter crib. 

(TS//SI) The usefulness of the Standard 
Grenades was increased when "G" discovered 
how to put them to exploiting German errors 
such as selecting wheel orders in a "Cilly" or non- 
random way. The Standard was also helpful in 
discovering why certain messages would not yield 
to the regular Bombe attack. The Grenades 
became a major way of exploring such "dud" mes- 
sages. 166 The Standard became a necessity at "G," 
and one was built into each of the second, 1944, 
models of the Bombes. 167 

( TS//SI) Once "G's" men had the opportunity, 
they began to build a series of additional 
Grenades; each provided an efficient solution to a 
particular Enigma problem, and each extended 
the power of the Bombes. 

CTO//0I) The Parallel Grenade allowed four 
short cribs to be used simultaneously to find the 
window settings. The Drag Grenade went further. 
It tested the four cribs, position by position, 
against a sixteen-letter cipher text. Its sister, the 
Polygrenade, was more powerful. It dragged the 

crib twenty-six positions at a time. The simpler 
Jones Dudbuster imitated the large paper cata- 
logs by dragging a common word such as "eins" 
through the text 


Q fOf/Olf The Sliding Grenades expanded the 
Bombe's power by handling those Enigmas with 
rotating reflector wheels. The Pluggable-Series 
Grenade was quite clever; it found the wheel 
order and ring setting for traffic produced using a 
double indicator. 169 

(T0//0I ) The most impressive of the 
Grenades was the Universal Plugboard. It was so 
flexible that it was used to explore many different 
ideas. Its main operational use was to try up to 
twenty-six cribs over thirteen letters - quite an 
imaginative use of plugs and wires. 

(TS//SI) Some of the Grenades were asked to 
solve more complex problems. To do so they used 
all of a Bombe or two Bombes lashed together. 
The Query found settings from the indicators on 
the messages. The Cilly automated the exploita- 
tion of nonrandom selection of part of a message 
setting. 170 

( TS//SI) Electronic devices were also used to 
make the Bombes more effective. Tube circuits 
were added to the first (N530) Bombes to com- 
pensate for weak menus that were producing too 
many stops. That Self-Detector was a set of the 
special four-in-one tubes attached to the original 
diagonal board. It suppressed stops that did not 
have a particular diagonal board connection: a 
letter connected to itself, that is, unsteckered. 
Statistical analyses had shown that it was very 
unlikely that a stop without such a connection 
would lead to the unraveling of an Enigma key. 171 

( T6//0I) The electronic Squelcher incorporat- 
ed a more general test to eliminate stops unlikely 
to produce key. It was a substitute for the original 
electronic amplifier system in the N530 Bombes. 
It was conceived when the worries about weak 
menus were intense. Once the Americans learned 


Page 113 


the tricks of menu building, they decided that 
only a few of their old Bombes needed to have the 
Squelchers. m 

(U) The success of Grenades and Squelcher 
circuits were only two of the indicators that 
Engstrom's group was beginning to overcome the 
chaos of 1942 and early 1943, By the fall of 1943 
things were going much, much better for 
Wenger's dreams for a permanent RAM program. 
In addition to the Bombes, OP-20-G finally began 
to receive the Gray-NCR and the Eastman 
machines. The first of the new Bush Comparators 
was put into operation in September. The new 
Comparator had a somewhat rocky career, how- 
ever. When it arrived in Washington, it had sev- 
eral flaws, including the incompatibility of its 
major components; incorrect specifications had 
been sent to the contractors. It also proved to be 
much slower than desired. But the complaints 
about the machine's failings were turned to "M's" 
advantage. The critics were assured that placing 
the next developments in the hands of the OP-20- 
GNCML group would prevent such mistakes. 173 
The failings ofEastman-Kodak's devices were 
also used as arguments for an expansion of "G's" 
own research and development. 174 All in all, by 
January 1944, OP-20-G's RAM group seemed 
vindicated and ready to return to the extension of 
the microfilm and digital electronic technologies. 
Some hoped there would be time to search for a 
general-purpose cryptanalytic machine, one that 
went beyond the Bush Comparator. 


1. (U) NSA Bombe File, U.S. Navy 0P-20-G/0P- 
20-2, Memo to Station X, "Decision regarding future E 
policy," app. May 1942. 

2. (U) Gordon Welchman, The Hut Six Story (New 
York: McGraw-Hill, 1982), 51. 

3.CFS7YM) On the Enigma, (TS//SI) NSA CCH 
Series XII Z, (S12OO8) Navy Dept., Office of Chief of 
Naval Operations, DNC (OP-20-G), RIP 425, "The 
American Attack on the German Naval Ciphers," 
October 1944. [sic]. 

4. CE8//6 fl NSA AHA 17580, Telford Taylor to 
Clarke and Corderman, "Early E. History," 10. 

5. (15//S1) NSA CCH Series XII Z, LeRoy H. 
Wheafley, "Cryptanalytic Machines in NSA" 30 May 
1953, and various years. {T S / f&Q . NSA CCH *'P" 
Collection Box CC067, RIP 608, CITS Paper TS-10/E- 
6, "Enigma Series Vol. 6, Duenna," CNC-OP-20, 
January 1946. NSA CCH Series IV.7.20, A P. Mahon, 
"The History of Hut Eight, 1939-1945," 12. (S) NSA 
AHA ACC 35701, "German Communications," 11 
October 1943. 

7. (U) Jean Stengers, "La guerre des messages 
codes, 1930—1945" LHisforie,3i(i98i): 19-31. Cipher 
A. Deavours, and Louis Kruh, Machine Cryptography 
and Modern Cryptanalysis (Dedham, Mass.: Artech 
House, 1985), 117. 

8.TSWWA AHA ACC 35701, "German Communi- 
cations," 11 October 1943. 

9.C TS//SI ) Andrew Hodges, Alan Turing: The 
Enigma (New York: Simon and Schuster, 1983), 175. 
Thomas Parrish, The Ultra Americans; The United 
States' Role in Breaking the Nazi Code (New York: 
Stein & Day, 1987), 49. Cipher A. Deavours, and Louis 
Kruh, Machine Cryptography and Modern 
Cryptanalysis (Dedham, Mass.: Artech House, 1985), 
117. W NSA CCH Series IV.7.20, A P, Mahon, "The 
History of Hut Eight, 1939-1945-" (TS//S1) NSA CCH 
Series XII Z, GCCS, "OP-20-G Contribution." 

10. (U) Wladyslaw Kozaczuk, Enigma (Frederick, 
Md.: University Publications of America, 1984). Jean 
Stengers, "Enigma, the French, the Poles and the 
British, 1931-1940," in Christopher Andrew and David 
Dilks (ed.), The Missing Dimension (London, 1984). 
Andrew Hodges, Alan Turing: The Enigma (New 
York: Simon and Schuster, 1983), 170-175- 

n. (TS//8I) NSA AHA 17580, Telford Taylor to 
Clarke and Corderman, "Early E. History." 

12. (D) Andrew Hodges, Alan Turing: The 
Enigma (New York: Simon and Schuster, 1983), 176. 
Some of the Scandinavian nations may have had spo- 
radic successes against Enigma and Fish, which they 
shared with Britain. But the true burden had to be car- 
ried by the British. Cryptologia, 12 (1988): 39> review 
ofW. M. Carlgren, Svensk Underrattelsetjanst 1939- 
45, (Stockholm, 1985). 

Page 114 



13. {¥&) NSA CCH Series IV.7.20, A. P. Mahon, 
"The History of Hut Eight, 1939-1945-" 

14. (U) Alan Hodges, Alan Turing: The Enigma 
(New York: Simon and Schuster, 1983). A.G. 
Denniston, "The Government Code and Cypher School 
between the Wars," Intelligence and National 
Security J (1986): 48-69. Christopher Andrew, Her 
Majesty's Secret Service: The Making of the British 
Intelligence Community (New York: Penguin, 1987). 

15. (U) Martin Campbell-Kelly, IC L:A Business 
and Technical History (Oxford: Clarendon Press, 
1989), 118. Gordon Welchman, The Hut Six Story: 
Breaking the Enigma Codes (New York: McGraw-Hill, 
1982), 295. Cipher A. Deavours and Louis Kruh, 
Machine Cryptography and Modern Cryptanalysis 
(Dedham, Mass.: Artech House, 1985), 119, 124. 

16. (U) David Kahn, Seizing the Enigma (Boston: 
Houghton Mifflin, 1991), 141. Andrew Hodges, Alan 
Turing: The Enigma (New York: Simon and Schuster, 
1983), 233. At one point Turing may have considered 
building a film machine for the method rather than 
using punched overlay sheets. It would have speeded 
the search or "coincidences" as did Bush's machine. 
But note that for Banburismus to be effective, captures 
of tables of some of the Enigma settings were vital. In 
1938, the British used another statistical test called 
SAGA It also appears to be the kind of approach used 
by Mrs. Driscoll in America. 

17. fFS.) NSA CCH Series IV.7.20, A P. Mahon, 
"The History of Hut Eight," 1939-1945-" 

18. (Y§) Scritching looked for contradictions pro- 
duced by assumption, concerning the settings of a 
cipher machine. It reduced the number of possibilities 
that have to be examined. CHS#SI) CCH Series XII Z, 
"Army- Navy, Descriptive Dictionary ofCryptologic 
Terms," ASA February, 1947. 

19. (W//SD NSA AHA ACC 13657, "G.C. & C.S. 
Naval S1GINT. Vol. Ill, German Cryptographic 
Systems and Their Solution." 

20. (W//SI) NSA AHA ACC 13657, "G.C. & C.S. 
Naval SIGINT. Vol. Ill, German Cryptographic 
Systems and Their Solution." 

21. CFSrVSI) NSA AHA ACC 13657, "G.C. & C.S. 
Naval SIGINT. Vol. Ill, German Cryptographic 
Svstems and Their Solution." 

22. (¥S) On other statistical methods, T^Si^NS A 
CCH Series IV, 7.20, A P. Mahon, "The History of Hut 
Eight, 1939-1945." 

23. (*FS) Avery readable and complete description 
of the British Bombe is found in NSA CCH Series XII 
Z, "The General Cryptanalytic Branch." 

24. (¥9-) NSA CCH Series IV.7.20, A P. Mahon, 
"The History of Hut Eight, 1939-1945 ■" 

25. (*S) NSA CCH Series IV.7.20, A P. Mahon, 
"The History of Hut Eight, 1939-1945," 28. 

26. (¥S//£L) Andrew Hodges, Alan Turing: The 
Enigma (New York: Simon and Schuster, 1983), 176. 
Gordon Welchman, The Hut Six Story: Breaking the 
Enigma Codes (New York: McGraw-Hill, 1982), 295. 
TrS7yStL NSA AHA ACC 13657, "G.C. & C.S. Naval 
SIGINT. Vol. Ill, German Cryptographic Systems and 
Their Solution." 

27. (U) Such commutator systems had been used 
in electrical timing instruments, but Keen did truly- 
creative work. 

28. C TO//0D NSA AHA ACC 13657, "G.C. & C.S. 
Naval SIGINT. Vol. Ill, German Cryptographic 
Systems and Their Solution." 

29. (¥S/SlQ Martin Campbell-Kelly, IC L: A 
Business and Technical History (Oxford, Clarendon 
Press, 1989), ll9."tW//SI) NSA AHA ACC 13657, "G.C. 
& C.S. Naval SIGINT. Vol. Ill, German Cryptographic 
Systems and Their Solution." 

30. (U) There are some different interpretations in 
Bradley F.Smith's, The Codebreaker's War: The 
Ultra-Magic-Deals (Novato: Presidio Press, 1993). 
Because of my evidence I remain convinced, for exam- 
ple, that the Americans took the lead in their Bombe 
program and that they intended to produce more than 
300 machines. See Smith, 247. 26. F. H. Hinsley, 
British Intelligence in the Second World War, Volume 
II (New York: Cambridge University Press, 1981), 747. 

31. (U) Jurgen Rohwer, The Critical Convoy Battle 
of March 1943 (London: Ian Allan, 1977), 240. Nigel 
West, GCHQ, the Secret Wireless War, 1900-86 
(London: Weidenfeld and Nicolson, 1986), 201, 210. 
F. H. Hinsley, British Intelligence in the Second World 
War, vol. Ill, Part I, (1984), 52. 

32. (S) NSA CCH Series XII Z, Robert L, Benson, 
"The Origin ofU.S. British Communications: 
Intelligence Cooperation (1940-41)," NSA Cryptologic 


Page 115 


Spectrum, 4 (Spring 1944): 5-8. (S) NSA CCH Series II 
Z, "History of GET (TUNNY) Research." 

33- (U) NARA RG457, SRH-361, "History of the 
Signal Security Agency, Volume Two, The General 
Cryptanalytic Problems," 017. Louis Kruh, "British- 
American Cryptanalytic Cooperation and an 
Unprecedented Admission by Winston Churchill," 
Cryptologia, 13(1989): 123-134- Nigel West, GCHQ, 
201. F. H. Hinsley, British Intelligence in the Second 
World War, Vol. 1 (London: Her Majesty's Stationery 
Office, 1979), 155- NSA RG457, SRH-197, "U. S.Navy 
Communications Intelligence, Organization, Liaison 
and Collaboration, 1941-45." But both Britain and the 
Dutch provided the Americans with intelligence and 
cryptanalytic help on Japanese systems earlier than 
previously thought. See Rear Admiral Edwin T. 
Layton, And I Was There: Pearl Harbor and Midway; 
Breaking the Secrets (New York: William Morrow, 
1985), 206. John W. M, Chapman, "Pearl Harbor: The 
Anglo-Australian Dimension," Intelligence and 
National Security I( 1989), 451-481. James 
Rusbridger and Eric Nave, Betrayal at Pearl Harbor 
(New York: Summit Books, 1991). 

34. (U) NARA RG457, SRH-141, "Papers from the 
Personal Files of Alfred McCormick, Part 2, March 4, 
1944," "Memorandum for General Bissel, Army-Navy 
Agreement Regarding Ultra." NARA RG457, SRH 152, 
"Historical Review of OP-20-G. 

35. "{Si NSA CCH Series XII Z, Robert L. Benson, 
"The Origin ofU.S. British Communications 
Intelligence Cooperation (1940-41)," NSA Cryptologic 
Spectrum 4 (Spring 1944), 5-8. 

36. TTSffSI) Lau ranee Safford, "Rhapsody in 
Purple," by Linda P.Tucker, Cryptologia, 6(1981), 
193-229, and 346-367. James Rusbridger and Eric 
Nave, Betrayal at Pearl Harbor (New York: Summit 
Books, 1991). NSASRH-391, "U.S.CTS#S1XNSACCH 
Series XII Z, "Washington E Traffic, Notes on 
Correspondence" circa February 1942. 

37. (U) David Kahn, Seizing the Enigma (Boston: 
Houghton-Mifflin, 1991), 235-6. NARA RG457, SRH- 
145, "Collection of Memoranda on Operations of SIS 
Intercept Activities and Dissemination 1942-1945," 
"Report of the Technical Mission to England" April 11, 
1941, 002-013. Greg Mellen (ed.), "Rhapsody in 
Purple: A New History of Pearl Harbor byDundas P. 

Tucker," Cryptologia, 6 (1981), 193-228. More bal- 
anced views are in Rear Admiral Edwin T. Layton, And 
I Was There: Pearl Harbor and Midway: Breaking 
the Secrets (New York: William Morrow, 1985) and 
Edward J.Drea, MacArthur's Ultra (University of 
Kansas Press, 1992). The Americans did not give 
everything to the British, however. NARA RG298, Box 
39, Memorandum ofR. W.Sylvester toL. Terman, 
April 11, 1942, tells Terman not to disclose any crypto- 
logical work to the British. And the Americans also 
kept their very important SIGCUM enciphering 
machine from the British during much of the war. See 
NARA RG457, RMA003, May 19,1944, Memorandum 
for Assistant Chief of Staff, G-2 from Office of the Chief 
Signal Officer. 

38. (U) NARA RG457, SRH-27O, "Army Navy FBI 
COMINT Agreements of 1942" by Robert L. Benson, 
and the very useful SRH-005, "Use of CX/MSS Ultra." 

39. $) NSA CCH Series TVV 10.1, "Report of the 
Technical Mission to England," 11-4-1941. 

40. (U) NARA RG457, SRH-361, "History of the 
Signal Security Agency, Volume Two, The General 
Cryptanalytic Problems," and SRH-145, "Collection of 
Memoranda on Operations of SIS Intercept Activities 
and Dissemination 1942-1945," 'Report of the 
Technical Mission to England' April 11, 1941, 002-013. 
TTSfm NSA CCH Series XII Z, (S12OO8) Navy Dept., 
Office of Chief of Naval Operations, DNC (OP-20-G), 
RIP 425, "The American Attack on the German Naval 
Ciphers," October 1944- [sfc]*CTS^ NSA CCH R 
Collection, Box CCO 66, OP-20-G-A, "American 
Cryptanalysis of German Naval Systems," 7 July 1944. 

4i.f8iLaurance Safford, "Rhapsody in Purple," by 
Dundas P.Tucker, Cryptologia, 6(1981), 193-229, and 
346-367. NARA, RG457, SRH-361, "History of the 
Signal Security Agency," 259, 261. There is some rea- 
son to believe that Britain did pass on the ways to crack 
the simpler 1940 German air force systems. NSA RAM 
File, "Report to J. N.Wenger, Capt. USN, Resume of 
the Dayton, Ohio Activity During World War II," and, 
J.T. Pendergrass, "Cryptanalytic Use of High-Speed 
Digital Computing Machines," Top Oecrot; 1946. "tS^ 
NSA CCH Series IV V 10.1, "Report of the Technical 
Mission to England," 11-4-1941. 

42. fFS^ NSA AHA ACC 35701 "History of the 
Bombe Project," 16 February 1946. "TS^ NSA CCH 

Page 116 



Series IV V io.i, "Report of the Technical Mission to 
England," 11-4-1941. (TS) NSACCH R Collection, Box 
CCo 66, OP-20-1 to OP-20-GY-A, 7 July 1944, 
"American Cryptanalysis of German Naval Systems." 

43. fFS) NSA CCH R Collection, Box CCO 66, OP- 
20-GY-A "American Cryptanalysis of German Naval 
Systems," 7 July 1944. 

44. f TO//01» NSA CCH Series XII Z, "Washington 
E Traffic, Notes on Correspondence" circa February 
1942. CFS) NSA AHAACC 35701 "History of the 
Bornbe Project," 16 February 1946." 

45. (¥&) NSA AHA ACC 35701 "History of the 
Bombe Project," 16 February 1946. 

46. (U) David Kahn, Seizing the Enigma (Boston: 
Houghton-Mifflin, 1991), 237. 

47. (¥S) NSA ARA ACC 35701 "History ofthe 
Bombe Project," 16 February 1946. (TS) NSA CCH 
Series IV V 10.6, Chief Signal Officer, "A Chronology of 
the Cooperation Between the SSA and the London 
Office of GCCS," 2 June 1946. 

48. (TS#SD NSA CCH Series XII Z, "Washington 
E Traffic, Notes on Correspondence" circa February 

49. (•§) An invaluable document on British-U. S. 
Army relationships, NARA RG457, SRH-005, "Use of 
CX/MSS Ultra," and of equal value, "History of 3-US," 
as in John Mendelsohn (ed.), Covert Warfare: 
Intelligence, Counterintelligence & Military 
Deception During the World War HEra (Garland, 
1989). (S) NSA CCH Series XII Z, Wenger toOP-20, 
October 1,1942, "Collaboration ofU. S.and British 
radio Intelligence organizations on Japanese and 
German projects." Robert L. Benson, A History of 
U. S. Communications Intelligence during World War 
II: Policy and Administration , Center for Cryptologic 

History 7 , NSA 1997- 

50. (U) For the complaints by the American army's 
SIS, see NARA RG457, SRH-361, "History ofthe 
Signal Security Agency" Vol. II. 249-275. 

51. (TS-) NSA AHA ACC 35701 "History ofthe 
Bombe Project," 16 February 1946. (TS) NSA CCH R 
Collection, Box CCO 66, OP-20-GY-A, "American 
Cryptanalysis of German Naval Systems," 7 July 1944. 
(S^NSA CCH series XII Z, Redman toOP-20, 2-28- 
1942, "British not cooperating." 

52. pPS) NSA AHA ACC 35701 "History ofthe 
Bombe Project," 16 February 1946.TTS7V«1^SA CCH 
Series XII Z,(Sl2008) Navy Dept., Office of Chief of 
Naval Operations, DNC (OP-20-G), RIP 425, "The 
American Attack on the German Naval Ciphers," 
October 1944- [sic] ffS) NSACCH R Collection, Box 
CCO 66, OP-20-GY-A, "American Cryptanalysis of 
German Naval Systems," 7 July 1944- 

53. (3S) NSA CCH Series XII Z, Travis toOP-20- 
G, 13 May 1942 - "Will Send Bombe to you in August 
or September." 

54. CRS^NSA CCH Series XII Z, Travis toOP-20- 
G, 13 May 1942 - "Will Send Bombe to you in August 
or September." 

55- ^SJ Andrew Hodges, Alan Turing: The 
Enigma (New York: Simon, and Schuster, 1983), 191. 
Laurance Safford, "Rhapsody in Purple," by Dundas P. 
Tucker, Cryptologia , 6 (1982), 216-17. NSA RAM File: 
OP-20-G to GC&CS, July 7, 1942; OP20-G to OP-20 
September 3, 1942; J. N. Wenger to OP-20-GM, 
August 6, 1942 "We wish to construct..."; and Wenger 
to Ely, August 5, 1942; Engstrom to Meader re Turing 
visit, January 5, 1943. F. H. Hinsley, British 
Intelligence in the Second World War, Vol. 1,56. 
Joseph Eachus, letter to the author, March 24, 1989. 
"History of 3-US," in John Mendelshon (ed.), Covert 
Warfare (Garland, 1989), 010-012. ffV/S Q On the 
continuing problems, it SIS, OKI) NSA CCH Series IV 
V10.6, Chief Signal Officer, "A Chronology ofthe 
Cooperation Between the SSA and the London Office 
of GCCS," 2 June 1946. On the August protest by"G," 
fES) NSA AHA ACC 35701 "History ofthe Bombe 
Project," 16 February i946ntS)kNSA CCH Series XII Z, 
Eachus to Wenger, 2 August 1942, "Full Wiring 
Diagram on Way." (SJ NSA CCH Series XII Z, Wenger 
to Op-20-GM, 6 August 1942, "Nature of E machine." 

56. (U) The group at Bletchley Park cracked the 
new four-wheel system by December 1942. But the 
break was not a pure cryptanalytic one. It depended 
upon the capture of documents and the continued fail- 
ure ofthe German system managers to follow basic 
security procedures. David Kahn, Seizing the Enigma 
(Boston: Houghton-Mifflin, 1991), 111. 

57. (U) Laurance Safford, "Rhapsody in Purple," 
by Dundas P.Tucker, Cryptologia, 6 (1982), 193-229, 
and 346-367. 


Page 117 


58. (U) Even Britain waited what now seems too 
long to organize its research. Ronald Clark, Tizard 
(Cambridge: MIT Press, 1965). 

59. (U) The Coast Guard, which had been in 
charge of decrypting rumrunner and other clandestine 
traffic for years, had as its chief cryptanalyst William 
Friedman's wife. The clandestine system the Coast 
Guard attacked must have been very simple compared 
to the Enigma. 

60. ffS^Sl) NSA CCH Series XII Z, (S12OO8) 
Navy Dept, Office of Chief of Naval Operations, DNC 
(OP-20-G), RIP 425, "The American Attack on the 
German Naval Ciphers," October 1944. [sic] 51, 53- 
e»).NSA CCH R Collection, Box CCO 66, OP20-GY-A, 
"American Cryptanalysis of German Naval Systems," 
7 July 1944. 

61. (U) NARA RG457, SRH-306, "OP-20-G 
Exploits and Commendations in World War II" 016. 
NSA RAM File, "Report to J. M. Wenger, Capt. USN, 
Resume of the Dayton, Ohio Activity During World 
War II," and J. T. Pendergrass, "Cryptanalytic Use of 
High-Speed Digital Computing Machines," Top Secret, 
1946. fPSl NSA CCH R Collection, Box CCO 66, OP- 
20-GY-A "American Cryptanalysis of German Naval 
Systems," 7 July 1944. 

62. CTSt^SI] NSA AHA ACC 13657, "G.C. &C.S. 
Naval SIGINT. Vol. Ill, German Cryptographic 
Systems and Their Solution." GC&CS learned of the 
possible introduction of the fourth wheel in May 1941 
but did not get Wynn-Williams to work until very late 
in the year. His work proceeded at a very slow pace. 
Doc Keen was not put to work to think about a high- 
speed Bombe for the four-wheel problem until late 
spring 1942. 

63. PE§i NSA CCH Series XII Z, Hut 6, 4 October 
1942, "Electronics for Bombe not Working." 

64. (U) Andrew Hodges, Alan Turing: The 
Enigma (New York: Simon and Schuster, 1983), 

65. (¥S#SJ) Keen built several new versions of his 
original Bombes. The new Mammoths and Jumbos 
handled the standard three-and four-wheel Enigma 
problems, while the Giants, Ogres and Twinns were, 
like some American variants, commutator-based 
machines for special Enigma variation problems. 
fF6#£D NSA AHA ACC 13657, "G.C. &C.S. Naval 

SIGINT. Vol. Ill, German Cryptographic Systems and 
Their Solution." Cr^SU NSA AHA ACC 17738, 
"E Operations of the GC&CS, list of machines," 
25 February 1945. 

66. TTS) Andrew Hodges, Alan Turing: The 
Enigma (New York: Simon and Schuster, 1983), 227. 
(8) NSA CCH Series XII Z,Eachus to Wenger, 18-9- 
1942, "British four wheel design not progressing." "fSSQ 
NSA CCH Series IV.7.20, A P. Mahon, 'The History of 
Hut Eight, 1939-1945. CS§) NSA CCH Series XIII Z, 
Hut 6,4 October 1942, "Electronics for Bombe not 
Working." t«8 NSA AHA ACC 35173, CNO, CITS TS- 
49, "A Posteriori Remarks on the Cryptanalytic 
Aspects of the Bulldozer," Navy Dept., Washington, 
September 1946. 

67. (TSt^SJQ NSA Bombe File, "It is desired to con- 
struct..." April 25, 1942, and August 5, 1942, Wenger to 
Ely. PHWSQ NSA CCH Series XII Z,(Sl2008) Navy 
Dept., Office of Chief of Naval Operations, DNC (OP- 
20-G), RIP 425, "The American Attack on the German 
Naval Ciphers," October 1944. [sic]. 

68. (WySTj NSA CCH Series XII Z, (Si2008) 
Navy Dept., Office of Chief of Naval Operations, DNC 
(OP-20-G), RIP 425, "The American Attack on the 
German Naval Ciphers," October 1944. [sic] 51, 53- 

69. flES) NSA CCH R Collection, Box CCO 66, OP- 
20-GY-A, "American Cryptanalysis of German Naval 
Systems," 7 July 1944- CFSJ NSA AHA 35529, 
Friedman to Corderman, 29 March, 1944, 
"Comparison of our "003" type of "Bombe" with the 
rotary type. 

70. Cm NSA AHA ACC 35701 "History of the 
Bombe Project," 16 February 1946. fFSWSI) Navy 
Dept., Office of Chief of Naval Operations, RIP 607, 
Enigma Series, volume 5., CIT TS-10-E-5, "Bombe 
Computations." CSSJ NSA CCH R Collection, Box CCO 
66, OP-20-GY-A, "American Cryptanalysis of German 
Naval Systems," 7 July 1944- (TStV&O NSA CCH P 
series Box CCO 67, RIP 607 shows that when "G" did 
finally incorporate a diagonal board it looked for hot- 
points; the British looked for coldpoints. 

71. fPSi NSA AHA ACC 35701 "History of the 
Bombe Project," 16 February 1946. ffS// SI) Navy 
Dept., Office of Chief of Naval Operations, RIP 607, 
Enigma Series, volume 5., CIT Ts-io-E-5, "Bombe 

Page 118 



72.TTS7VSi). Navy Dept., Office of Chief of Naval 
Operations, RIP 607, Enigma Series, volume 5-, C1T 
Ts-io-E-5, "Bombe Computations." 

73. rPW&e-Navy Dept., Office of Chief of Naval 
Operations, RIP 607, Enigma Series, volume 5., CIT 
Ts-io-E-5, "Bombe Computations." (T0//M J NSA 
CCH Series XII Z (S12008) Navy Dept., Office of Chief 
of Naval Operations, DNC (OP-20-G), RIP 425, "The 
American Attack on the German Naval Ciphers," 
October 1944- [sic] 51-3- f«0 NSA CCH R Collection, 
Box CCO 66, OP-20-GY-A "American Cryptanalysis 
of German Naval Systems," 7 July 1944. 

74. (U) NSA RAM File, "Decision Regarding 
Future E Policy," app May 1942, and "For GC&CS" 
August 5, 1942. 

75. (U) NSA RAM File, app. May, 1942, 0P-20-G 
to Station X, "Latest Thoughts on Electronic 

76. (U) Much of the information on GC&CS's 
actions and intention remains in closed archives, and 
some informed historical guesses have to be made. 
One of those is that GC&CS's failure to send requested 
documentation was as much the result of the slow pace 
of its four-wheel Bombe program and a desire to keep 
face as it was the result of a desire to monopolize all the 
"E" work. 

77. CFS) NSA AHA ACC 35701 "History of the 
Bombe Project," 16 February 1946, 4. 

78. (SJ Britain sent more assurances to the United 
States in August, but they were too late to stop "G" 
from going ahead with its Bombe program, fij) NSA 
CCH Series XIII Z, Eachus toWenger, 2 August 1942, 
"Full Wiring Diagram on Way." (Si NSA CCH Series 
XII Z, Wenger toOp-20-GM, 6 August 1942, "Nature 
of E machine." 

79- (U) NSA RAM File, OP-20-G to GC&CS, July 7, 
1942, "Eachus &Ely," and August 5,1942, "Send 
Wiring Diagram." NARA RG457, SRH-306, "OP-20-G 
Exploits and Commendations in World War II," 23, 
"Shark was not all." Andrew Hodges, Alan Turing: The 
Enigma (New York: Simon and Schuster, 1983), 236. 
F. H. Hinsley, British Intelligence in the Second World 
War, Volume II (New York: Cambridge University 
Press, 1981), 55-57- 

80. (TS//SQ. The Robinsons were independent 
creations but were much like the Bush comparator 

machines. However, they were built to handle the 
codes used on teletype systems, and they were not 
strict embodiments of the IC method but of Turing's 
new Delta method, which searched for statistical bias- 
es in the distribution of binary values. However, Delta 
shared much with the IC approach. (TS//&O (Laconic, 
Nocon) NSA CCH Series XII Z, Glenn F. Stahly, "Fifty 
Years of Mathematical Cryptanalysis," (T3//GI NSA 
CCH Series XII Z, H. H, Campaigne, "Reading 
TUNNY," NSA Technical Journal (Fall 1962). 
T T 3//M ) NSA CCH Local Archive, "Army-Navy 
Descriptive Dictionary ofCryptologic Terms," Army 
Security Agency, February 1947. 

81. TTV/^B" Letters from Howard Campaigne to 
Brian Randell circa 1975. Thomas H. Flowers, "The 
Design of Colossus," Annals of the History of 
Computings (July 1983), 224. I.J. Good, "Early 
Work on Computers at Bletchley," Cryptologia 3 
(1979), 65-77- (TS77«t) NSA CCH Series XII Z, H. H. 
Campaigne, "Reading TUNNY," NSA Technical 
Journal (Fall 1962). 

82. (T9//SI). The Robinsons, also called "bed- 
steads," became more and more powerful and rapid. 
Some models used vacuum tubes, and by the war's 
end, one of the Robinsons used four tapes.'T^NSA 
CCH Series XII Z,GCHQ, "Machine Solution of 
TUNNY Traffic (Robinson)," 22 August 1943-TTS7V«^ 
NSA CCH Local Archive, "Army-Navy Descriptive 
Dictionary ofCryptologic Terms," Army Security 
Agency, February 1947. 

83. (U) NSA CCH XI K, S. Snyder, Box, II, B. 
Randell, "The Colossus," June 1976. rTS7VSIl.NSA 
CCH Series XII Z, "Theory of Rectangles: Photostat of 
British Paper describing breaking of the TUNNY 
Machine by Means of Rectangles," 4 September 1944. 
ffS//SQ NSA CCH Series XII Z,"Fish Notes," 
December 1944- (TS//SI) NSA CCH Series XII Z, 
"Colossus, Instructions and procedures used in setting 
FISH messages on the Colossus," 14 December 1944. 

84. (T577<SJ) NSA CCH Series XII Z, "Colossus, 
Instructions and procedures used in setting FISH mes- 
sages on the Colossus," 14 December 1944. 

85. (U) Thomas H. Flowers, "The Design of 
Colossus," Annals of the History of Computing 5 (July 
1983): 240. Allen W.M. Coombs, "The Making of 
Colossus" Annals of the History of Computing 5 


Page 119 


(1983): 253-259. W.W. Chandler, "The Installation 
and Maintenance of Colossus," Annak of the History 
of Computing 5 (1983): 260, Andrew Hodges, Alan 
Turing: The Enigma (New York; Simon and Schuster, 
1983), 268, Although their documentation is still clas- 
sified, there were additional electronic machines in 
GC&CS's arsenal by the end of the war. Apparently, 
they went beyond Colossus. 

86. (U) NARA RG457, SRH-349, "Achievements 
of the SSA in World War II," and SRH-361, "History- of 
the Signal Security Agency." 

87. (U) NSA RAM File, OP-20-G to GC&CS, app. 
May, 1942, 'Future E Policy,' and Wenger to Ely, 
August 5, 1942. 

88. (U) Ronald Lewin, The American Magic (New 
York: Farrar, Straus, Giroux, 1982), 85. Rear Admiral 
Edwin T. Layton, etal., And I Was There: Pearl 
Harbor and Midway: Breaking the Secrets (William 
Morrow & Co., Inc., 1985), 95. W, J. Holmes, Double- 
Edged Secrets (Annapolis: Naval Institute Press, 
1979). One story is that Washington found different 
additives from their JN-25 analyses and thought that 
Hawaii was far off base regarding the recovered code 

89. (U) NSA RAM File, August 5, 1942, Wenger to 
Ely. NSA RAM File, Part II of Report to J. N. Wenger, 
Capt. USN, "Resume of the Dayton, Ohio Activity 
During World War II." Hagley Museum and Library, 
Accession 1825, Honeywell v Sperry-Rand, Trial 
Records, August 19, 1942, Desch to Engineering 
Department, "Special Switch," and September 18, 
1942, "Change in Specifications." 

90. (U) NSA RAM File, Part II. of Report to J. N. 
Wenger, Capt. USN, "Resume of the Dayton, Ohio 
Activity During World War II." 

91. (U) A simple electronic American Bombe 
would have sixty-four fixed double wheels calling for at 
least 7,000 tubes, probably twice that. Amplifiers and 
control electronics would probably have called for 
another 10,000 or so. A universal machine, in which 
all wheels could be automatically set, would have 
needed close to 100,000 tubes. Correspondence with 
Joseph Eachus. NSA RAM File, Wenger to GC&CS, 
September 4, 1942, "Electronic Device." 

92. (U) F. H. Hinsley, Biitish Intelligence in the 
Second World War, Volume III (New York: 

Cambridge University Press, 1981), 56. For the inde- 
pendent research agreement, interview with Joseph 
Eachus. The situation with the American army may 
have been even more critical. It was a pure consumer 
of Ultra and much other intelligence during the North 
African invasion and as late as February 1943, it was 
dependent on GC&CS for German signal intelligence. 
NARA RG457, SRH-364, "History of the Signal 
Security Agency, Volume One Parts 1 and 2, 1939- 
1945- A Declaration of Independence." 

93-"TS)sNSA CCH Series XII Z, Eachus to Wenger, 
18-9-1942, "British four-wheel design not progress- 

94. e^ NSA AHA ACC 35701 "History of the 
Bombe Project," 16 February 1946, 4-*t*S^NSA CCH 
Series XII Z, Engstrom to Desch, September 23 1942, 
"Your Bombe plan approved." 

95. (U) NSA RAM File, September 3,1942, 
Wenger toOP-20, "Cryptanalysis of the German 
(Enigma) Machine." F. H. Hinsley, British Intelligence 
in the Second World War, Volume II (New York-, 
Cambridge University Press, 1981), 57- Hagley 
Museum and Library, Accession 1825, Honeywell v 
Sperry-Rand, Trial Records, September 18, 1942, 
Desch to Engineering Department. Compare the 
Americans' promised production rate with Keen's out- 
put. Martin Campbell-Kelly, IC L:A Business and 
Technical History, (Oxford: Clarendon Press, 1989), 
118-119. If the United States had built the 350 
machines rather than the 100 in the first batch, the 
cost would have been a staggering $16,000,000 based 
on the $45,000 per machine for the production run. 

96.^1 NSA CCH, Series XII Z, Joseph Desch to 
OP-20-G, "Memo of present plans for an electro- 
mechanical analytical machine," 15 September 1942. 

97. ^ NSA CCH Series XII Z, Report of 
Dr. Turing, "Visit to National Cash Register 
Corporation of Dayton, Ohio," circa Dec. 1942. 

98. dm NSA CCH Series IV.7.20, A. P. Mahon, 
'The History of Hut Eight, 1939-1945," 6& states that 
the later versions of the British high-speed Bombe had 
fast wheels which moved at 10,000 revolutions per 
minute. But this figure may be a typographical error. 
Most other sources give a much lower rpm for the 
British four-wheel Bombe. See (Y&) NSA CCH Series 
XII Z, Hut 6, "Electronics for Bombe not Working," 

Page 120 



which in late 1942 gives a projected speed of 2,000 
rpm. Also see ffSJ NSA CCH Series XII Z,Joan 
Murray, "A Personal Contribution to the Bombe 
Story," NSA Technical Journal, 20 (Fall 1974): 41- 

99. CS) NSA CCH, Series XII Z.Joseph Desch to 
OP-20-G, "Memo of present plans for an electro- 
mechanical analytical machine," 15 September 1942. 

100. CSi. NSA CCH, Series XII Z, Joseph Desch to 
OP-20-G, "Memo of present plans for an electro- 
mechanical analytical machine," 15 September 1942, 


101. fS) A full diagonal board would be a 26 x 26 
matrix, which would need 676 primary tubes and a 
score of complementary electronic components. 
CS} NSA CCH, Series XII Z, Joseph Desch to OP-20-G, 
"Memo of present plans for an electromechanical ana- 
lytical machine," 15 September 1942, 9. 

102. CS) NSA CCH, Series XII Z, Joseph Desch to 
OP-20-G, "Memo of present plans for an electro- 
mechanical analytical machine," 15 September 1942, 1. 

103. f¥ S//& t) Navy Dept., Office of Chief of Naval 
Operations RIP 607, Enigma Series, volume 5, CIT Ts- 
10-E-5, "Bombe Computations." 

104. (f fi/ZS t) NSA CCH Series XII Z, 
"Cryptanalytic Report #2: The "Yellow Machine," 50- 
52. ( T 5//3I) Navy Dept., Office of Chief of Naval 
Operations, RIP 607, Enigma Series, volume 5, CI'F 
Ts-io-E-5, "Bombe Computations." (TS//S1) NSA 
AHA 16331, "6812th Signal Security Detachment 
(PROV) Apo 413 Army," 15 June 1945. 

105. GFS-) NSA CCH Series XII Z, Engstrom to 
Desch, September 23, 1942, "Your Bombe plan 
approved," $} NSA CCH, Series XII Z, Joseph Desch 
toOP-20-G, "Memo of present plans for an electro- 
mechanical analytical machine," September 15, 1942, 

106. ( T S //SI) The men at "G" were not able to 
intercept and interpret strong menus for quite some 
time, they sent very weak menus to the Bombes 
throughout much of 1942 and 1943, They realized they 
were dependent upon the British for the types of 
menus that made the Bombes useful by having very 
few "stories" produced per run. (fSi NSA AHA ACC 
35701 "History of the Bombe Project," 16 February 
1946. ffSZ/SH NSA CCH Series XII Z, GCCS, "OP-20- 
G Contribution." (TW&Q Navy Dept., Office of Chief 

of Naval Operations, RIP 607, Enigma Series, volume 
5., CIT TS-10-E-5, "Bombe Computations." TS^NSA 
CCH Series XII Z, "Cryptanalytic Report #2: The 
Yellow Machine," 50-52. 

107. fFS} NSA CCH Series XII Z, Engstrom to 
Desch, September 23, 1942, "Your Bombe plan 
approved." fS)i NSA CCH, Series XII Z.Joseph Desch 
toOP-20-G, "Memo of present plans for an electro- 
mechanical analytical machine," September 15, 1942, 

108. tS) NSA CCH, Series XII Z, Memorandum for 
OP-20-G, "Collaboration ofU. S.and British Radio 
Intelligence Organizations on Japanese and German 
Projects," J. N. Wenger, October 1, 1942. 

109.1S).NSA CCH, Series XII Z, Memorandum 
forOP-20-G, "Collaboration ofU. S.and British Radio 
Intelligence Organizations on Japanese and German 
Projects," J. N. Wenger, October 1, 1942. 

110. (U) Hagley Museum and Library, Accession 
1825, Honeywell v Sperry-Rand , Trial Records, 
September 18, 1942, Desch to Engineering 
Department. F. H. Hinsley, British Intelligence in the 
Second World War, Volume II (New York: Cambridge 
University Press, 1981), 56. NARA RG457, SRH 361, 
"History of the Signal Security Agency," 274. NSA 
RAM File: OP-20-G to GC&CS July 7, 1942; OP-20-G 
toOP-20 September 3,1942; J.N. Wenger toOP-20- 
GM, August 6, 1942; Wenger to Ely, August 5, 1942; 
and "Engstrom toMeader re Turing visit," January 1, 

111. CSJ NSA CCH, Series XII Z, Memorandum for 
OP-20-G, "Collaboration ofU. S.and British Radio 
Intelligence Organizations on Japanese and German 
Projects," J. N. Wenger, October 1, 1942. 

112. (U) Of course, the Americans did not suspect 
that Britain had its own Russian informant who was 
telling all to Stalin. Christopher Andrew and Oleg 
Gordievsky, KGB: The Inside Story of Its Foreign 
Operations from Lenin to Gorbachev (New York: 
Harper Collins, 1990), 304. 

113. (U) NARA RG457, SRH-306, "0P-20-G 
Exploits and Commendations in World War H," 19. F. 
H. Hinsley, British Intelligence in the Second World 
War, Volume II (New York: Cambridge University 
Press, 1981), 57. Andrew Hodges, Alan Turing: The 
Enigma (New York: Simon and Schuster, 1983), 243. 


Page 121 


tS^NSA CCH Series XII Z, Wenger to OP-20, October 
1,1942, "Collaboration ofU. S.and British radio 
Intelligence organizations on Japanese and German 

114. (U) Very useful on the question of the U.S. 
Army's Ultra struggle is "Origins, Functions & 
Problems of the Special Branch, M.I. S.," in John 
Mendelshon (ed.), Covert Warfare: Intelligence, 
Counterintelligence and Military Deception During 
the World War II Era (Garland, 1989). 

115-TTS^NSA CCH Series IV.7.20, A. P. Mahon, 
The History of Hut Eight, 1939-1945," 90. 

116. (TS//SI) NSA CCH Series XII Z, (S 12008) 
Navy- Dept., Office of Chief of Naval Operations, DNC 
(OP-20-G), RIP 425, "The American Attack on the 
German Naval Ciphers," October 1944 [sic], cites 
October 1942 as the date of arrival of the first M8, but 
other sources claim it arrived in spring 1943. On the 
M9, interview with Philip J. Bochicchio, 14 September 

117. (J9ffS$ NSA CCH Series XII Z, (S 12008) 
Navy Dept., Office of Chief of Naval Operations, DNC 
(0P-20-G), RIP 425, "The American Attack on the 
German Naval Ciphers," October 1944. [sic] (TStYSI) 
NSA CCH "P" Series, Box CCo67, CITS, "E" Series, 
Vol. E-7. 

118. ff//S i). Andrew Hodges, Alan Turing; The 
Enigma (New York: Simon and Schuster, 1983), 236. 
NSA RAM File, January 5, 1943, "Report on Turing 
Visit to Dayton." (S) NSA CCH Series XII Z, Dr. Turing 
ofG.C. &C.S., "Visit to National Cash Register 
Corporation," December, 1942. (TS//S l r)- NSA CCH 
Series XII Z, OP-20-G "Memoranda on Bombe and 
the relationship of the U. S. and U. K.," circa 1943. 

119. tS) Turing remarked about Desch's apparent 
decision not to use some features already planned for 
the new British four-wheel Bombe. {S}- NSA CCH 
Series XII Z, Report of Dr. Turing, "Visit to National 
Cash Register Corporation of Dayton, Ohio," circa Dec. 

120. (SJ NSA CCH Series XII Z, Eachus to Wenger, 
18 September 1942, "British four wheel design not pro- 
gressing.'" CSi NSA CCH Series XIU Z, Report of Dr. 
Turing, "Visit to National Cash Register Corporation of 
Dayton, Ohio," circa Dec. 1942. 

121. (¥SJ. On British comments besides those of 
Turing, NSA CCH, shinn box, "Desch." 

122. (¥§}. NSA CCH Series XII Z, Engstrom to 
Desch, September 23, 1942, "Your Bombe plan 

123. r8i NSA RAM File,, January 5, 1943, "Report 
on Turing visit to Dayton"; January 20, Engstrom to 
Meader, "Change Bombe Design"; and March 17, 1943 
Prototypes constructed.. fSJ NSA CCH Series XII Z, 
Report of Dr. Turing, 'Visit to National Cash Register 
Corporation of Dayton, Ohio," circa. December 1942. 
(£) NSA CCH Series XII Z,Dr. Turing ofG.C.& C.S. 
"Visit to National Cash Register Corporation," 
December 1942. 

124. tSi On the gearing, interview with Philip J. 
Bochicchio, 6-14-94. 

125. (SiNSA CCH Series XII Z, Dr. Turing of G.C. 
& C. S., "Visit to National Cash Register Corporation," 
December 1942. 

126. CTO//0*) NSA CCH Series XII Z.OP-20-G 
"Memoranda on Bombe and the relationship of the 
U.S. and U.K.," circa 1943. 

127. (U) Interview with Philip J. Bochicchio, 6-14- 

128. (U) Interview with Philip J. Bochicchio 6-14- 


129. (*FS). Interview with Philip J. Bochicchio, 6- 
14-94. NSA, CCH Series XII Z,CNO, CITP 88, 
"Technical and Theoretical Report of N-530 BOMBE, 
Navy Dept., Washington, D.C., September, 1946. 

130. rT^SQ-NSA RAM File, December 11, 1942, 
Horn to Robinson, "Procurement of Materials for 
7892," and December 28, 1942, CNO to Deeds, "Help 
needed on special project at NCR." (T9ff&i± NSA CCH 
Series XII Z, OP-20-G "Memoranda on Bombe and 
the relationship of the U. S. and U. K.," circa 1943. 

131. (U) NSA RAM File, January 5, 1943, Wenger 
to Meader, "Turing Visit," Andrew Hodges, Alan 
Turing; The Enigma (New York: Simon and Schuster, 

1983), 236- 

132. T^S) NSA AHA ACC 35701 "History of the 
Bombe Project," 16 February 1946. 

133. (U) NSA NCML Message File, for example, 
June 4, 1943, "Use most Adam-Eve time for real tests." 

134. (U) NSA NCML-CSAW Message file, May 24, 
1943, "DC has sent test problems for prototypes," and 

Page 122 



May 29, 1943, "Adam Eve have serious technical prob- 
lems." The first British four-wheel Bombe was com- 
pleted in June 1943. F. H. Hinsley, British Intelligence 
in the Second World War, Volume II (New York: 
Cambridge University Press, 1981), 748. 

135. (U) NSA NCML-CSAW Message File, May 29, 
\ 1943, "Shorts and opens," and May 31, 1943, "Can use 

part of machine only." 

136. t ra//S 9 NSA NCML-CSAW Message File, 
May 20, 1943,'Redman Visit," and May 29, 1943 
"Adam and Eve problems." F. H. Hinsley, British 
Intelligence in the Second World War, Volume II 
(New York: Cambridge University Press, 1981), 748. 
Andrew Hodges, Alan Turing: The Enigma (New 
York: Simon and Schuster, 1983), 267. W. W. 
Chandler, "The Installation and Maintenance of 
Colossus," Annals of the History of Computing 
5 (1983): 261. ffSJ On the date of England's first four- 
wheel Bombe, ff9//S*NSA CCH Series XII Z, GCCS, 
OP-20-G Contribution. 

137. (U) NSA NCML-CSAW Message File, June 
14,1943, "Cain and Abel." 

138. (U) The first "formal" production model was 
turned over to the navy at Dayton for testing on July 4, 
1943, and its first test run was on July 23, 1943. Cain 
and Abel were off the line in the first week of July but 
were not ready for final testing until the last week of 
the month. Another complication was the delayed 
completion of the new building in Washington. Desch 
did make some minor changes after Engstrom's 
request. NSA NCML-CSAW Message File, July 6, 
1943, July 26,1943, and July 23, 1943, "Status of 

139. (U) NSA NCML-CSAW Message File, July 26, 
1943, Dayton to Washington 'Bombes may not work.' 

140. (U) NSA NCML-CSAW Message File, July 29, 
1943, Desch to Engstrom, July 29, 1943, "Fast wheel 
running too hot, bombe may not work." 

141. (U) Interview with Philip J. Bochicchio, 6-14- 

142. (TS) NSA AHA ACC 35173, CNO, CITS TS-49, 
"A Posteriori Remarks on the Cryptanalytic Aspects of 
the Bulldozer," Navy Dept., Washington, September 

143. (TS.) NSA CCH Series XII Z, Hut 6, 4 October 
1942, 'Electronics for Bombe not Working.' fSSJNSA 

CCH Series XII Z, Engstrom to Desch, September 23, 
1942, "Your Bombe plan approved." 

144. (IS). NSA CCH Series XII Z, Joan Murray, 
"A Personal Contribution to the Bombe Story," NSA 
Technical Journal, 20 (Fall 1974): 41- 

145. 4S)-NSA CCH Series XII Z, Joseph Desch, 
"Plan for the American Bombe," September 15, 1942. 
(TO//OI) NSA CCH Series XII Z (S-2568) "Tentative 
Brief Descriptions of Cryptanalytic Equipment for 
Enigma Problems," Circa 1945. 

146. T T3//0I) NSA CCH Series XII Z (S-2568) 
"Tentative Brief Descriptions of Cryptanalytic' 
Equipment for Enigma Problems," Circa 1945. 

147. (TO//SI) NSA CCH Series XII Z, (S12008) 
Navy Dept., Office of Chief of Naval Operations, DNC 
(OP-20-G), RIP 425, "The American Attack on the 
German Naval Ciphers," October 1944. [sic] tS) NSA 
CCH Series XII Z, "Cryptanalytic Report #2: The 
Yellow Machine," 50-52. 

148. (TO//OI) Navy Dept., Office of Chief of Naval 
Operations, RIP 607, Enigma Series, volume 5., CIT 
Ts-io-E-5, "Bombe Computations," 5-14. fS)- NSA, 
CCH Series XII Z, "Cryptanalytic Report #2: The 
Yellow Machine," 12, 16. ( TQ//OI) NSA AHA 16331, 
"6812th Signal Security Detachment (PROV ) Apo 413 
Army," 15 June 1945- l"S*l NSA, CCH Series II Z, 
CNO, C1TP 88, "Technical and Theoretical Report of 
N-530 BOMBE," Navy Dept., Washington, D.C., 
September 1946. 

149- (Si NSA CCH Series XII Z, "Cryptanalytic 
Report #2 The Yellow Machine," 3, 54. 

150. ffS) NSA CCH Series.V.7.20, A. P. Mahon, 
"The History of Hut Eight, 1939-1945." 

151. (U) NSA NCML-CSAW Message File, 
November 12, 1943, "85 Bombes in D.C." 

152. (U) The first Bombe contract was terminated 
on December 1, 1943, but some of the first models con- 
tinued to trickle into Washington as late as Summer 
1944. NSA RAM File, History of OP-20-G /NCML/4e, 
June, 1944, "n530 bombes." 

153- (U) NSA NCML-CSAW Message File, 
November 29, 1943, 'Bombe shipment from Dayton.' 

154- (U) H. F. Hinsley, et al., British Intelligence in 
the Second World War .Volume II (New York: 
Cambridge University Press, 1981), 57-8 and 752. 

top oconcT//oo Mi NT//ra:L to uca, auc, cam, g b r amd nau/xi 

Page 123 


Andrew Hodges, Alan Turing: The Enigma (New 
York: Simon and Schuster, 1983), 262. 

155. (T3//3I) NSA CCH Series XII Z,(Si2008) 
Navy Dept., Office of Chief of Naval Operations, DNC 
(OP-20-G), RIP 425, "The American Attack on the 
German Naval Ciphers," October 1944, [sic] 23. 

156. (T9} NSA CCH Series IV.7.20, A. P. Mahon, 
The History of Hut Eight, 1939-1945." 

157. (U) H. F. Hinsley, et al., British Intelligence in 
the Second World War, Volume II (New York: 
Cambridge University Press, 1981), 752. NARA 
RG457, SRH-141, Papers from the Personal Files of 
Alfred McCormick, Part 2," March 4, 1944, 
"Memorandum for General Bissel, Army-Navy 
Agreement Regarding Ultra" Thomas Parrish, Tlie 
Ultra Americans: The United States' Role in Breaking 
the Nazi Code (New York: Stein & Day, 1987), 79- 
NSA RAM File, February 21, 1944, W. A, Wright to OP- 
20-G,"Comparison of Army and Navy Enigma 

i58.-f¥SH<SA CCH R Collection, Box CCO 66, OP- 
20-GY-A, "American Cryptanalysis of German Naval 
Systems," 7 July 1944- (Tfl//CI) NSA AHA ACC 13657, 
"G.C. &C.S. Naval SIGINT. Vol. Ill, German 
Cryptographic Systems and Their Solution." 

159. (Wf NSA AHA ACC 35701 "History of the 
Bombe Project," 16 February 1946, 9-TS^NSA CCH 
Series XII Z, 21-2-1944, OF-20-G to British Admiralty 
Delegation, "U.S. Will Increase Efforts. " (TS//SI) NSA 
CCH Series XII Z, Alexander toLt. Church, 24 March 
1944, "British will emphasize building three-wheel 

160. <TS//S1) NSA CCH Series XII Z, LeRoy H. 
Wheatley, "Cryptanalytic Machines in NSA," 30 May 
1953, and various years. (TS//BI) NSA CCH Series XII 
Z, copies of various MAC Outlines, circa 1953, 

161. (U) NSA NCML-CSAW Message File, July 20, 
1943, "New procedure, grenade," and August 14, 1943, 
"Progress on grenade." The grenades were also used 
on the four-wheel problems when it was thought that 
one of the wheel settings was known. 

162. (TS//SI) NSA CCH Series XII Z, LeRoy H. 
Wheatley, "Cryptanalytic Machines in NSA," 30 May 
1953, and various years. F TO//03 NSA CCH Series XII 
Z, copies of various MAC Outlines, circa 1953. 

163. ffS77«^-NSA CCH Series XII Z(S-2568) 
"Tentative Brief Descriptions of Cryptanalytic 
Equipment for Enigma Problems," circa 1945. 

164. (U) NSA NCML-CSAW Message File, July 
20, 1943, "New procedure, grenade," and August 14, 
1943, "Progress on grenade." (€) NSA CCH Series XII 
Z, Descriptions of NSA Early SPDs and Computers, as 
compiled from various NSA sources. 

165. flfl// S3 NSA CCH "P" Series, Box CC006, 
CNO CITS TA-10-E-1, Volume i,The Click Process, 
January 1946. She pointed to the use of the message 
indicators as the "crib," but other cribs could be used 
by the Grenades. 

i66.t€i.NSA CCH Series XII Z, Descriptions of 
NSA Early SPDs and Computers, as compiled from 
various NSA sources. 

i67- - ( TS//S» NSA CCH Series XII Z, LeRoy H. 
Wheatley, "Cryptanalytic Machines in NSA" 30 May 
1953, and various years. (TS//SI) NSA CCH Series XIT 
Z, copies of various MAC Outlines, circa 1953. 

i68 r(T0//ai)- NSA CCH Series XII Z,'Le.Roy H. 
Wheadey, "Cryptanalytic Machines in NSA" 30 May 
1953 and various years. {TS7ySJQ.NSA CCH Series XII 
Z, copies of various MAC Outlines, circa 1953. 

169. {€)-NSA CCN Series XII Z, Descriptions of 
NSA Early SPDs and Computers, as compiled from 
various NSA sources. 

l70. -(TD//ei) M SA CC* Series Xll Z, LeRoy H. 
Wheatley, "Cryptanalytic Machines in NSA," 30 May 
1953 and various years."TT3//6l).NSA CCH Series XII 
Z, copies of various MAC Outlines, circa 1953. 

171. -&& (S2568) NSA CCH Series XII Z, 
"Tentative Brief Description of General Analytic 
Equipment for Enigma Problems," 26 March 1945. 

172. £¥« (S2568) NSA CCH Series XII Z, 
"Tentative Brief Description of General Analytic 
Equipment for Enigma Problems," 26 March 1945- "A 
Time of Triumph." 

173. (U) NSA RAM File, CNO, U.S. Naval 
Communications, "Brief Descriptions of RAM 
Equipment," Washington, D.C., October 1947. NSA 
RAM File, Report of R I. Meader, Captain USNR to J. 
H.Wenger, Captain, USN, "14 Days Training Duty, 
Report of," January 21, 1949. 

Page 124 



174. (U) NSA RAM File, CNO, U, S, Naval 
Communications, CITP TP-33 "Overhaul of Hypo #1," 
Washington, D.C., June 1945- 



This page intentionally left blank 



Chapter 5 
(U) A Search for Other "Bombes" 

(U) The arrival of the United States Navy's 
Bombes in Washington in autumn 1943 allowed 
OP-20-GM to turn its attention to Japan. It also 
gave some of its men time to think of advancing 
beyond electromechanics. But the tenuous con- 
trol over the Enigma systems and the challenges 
of the very stubborn Japanese codes and ciphers 
meant that electromechanics and the Bombes 
continued to demand much of the energies of the 

(U) OP-20-GM explored many electronic and 
photo-optical possibilities during the last two 
years of the war as it attempted to conquer 
Japan's systems, and as it responded to Britain's 
cries for help to fight changes in Germany's codes 
and ciphers. In a few instances "M" was able to go 
beyond the technology of the Bombes, but in 
most cases it had to relegate electronics and 
advanced film-based processing to small 
exploratory projects. Only when there was a com- 
bination of an inescapable demand for ultra-high 
speeds and a possibility of coaxing electronics 
into behaving would "G" allow its engineers to try 
to turn their electronic dreams into hardware. 

(U) The army's SIS also had to drop its ambi- 
tious early plans for advanced electronic devices. 
Like the navy's cryptanalysts, its men had to turn 
to quick and rather clumsy solutions during the 
first years of the war. 

(U) Meanwhile, the Army 

(U) In late 1942, while OP-20-G's cryptana- 
lysts were establishing their place in European 
communications intelligence, the American 
army's codebreakers struggled to gain just a 
foothold. Unlike the navy, the army was not 
involved in European-related action until well 

after the outbreak of the war. It had a more diffi- 
cult time than the navy in intercepting enemy 
messages, and the British were much less in need 
of its cooperation. 1 

(T0//0I) The British had begun to share their 
knowledge of German and Italian diplomatic traf- 
fic before the war, but they were more than reluc- 
tant to allow the American army a role in the 
German army and air force systems. 2 At first, the 
Signal Corps and the SIS were not worried about 
their inability to read the German military traffic. 
Just before Pearl Harbor they indicated they were 
not interested in working on the army and air 
force problems. But when troops were committed 
to North Africa, attitudes changed dramatically. 
The Americans realized the shortcomings of 
depending on intelligence supplied by another 
nation. The SIS wanted its own control over 
Enigma, but it had few capabilities. 3 

(U) Founded to replace Herbert Yardley's 
infamous Black Chamber in the late 1920s, the 
army's Signal Intelligence Service (SIS) began 
with what Joseph Wenger yearned for, a core of 
young and talented civilian mathematicians. 
Under William F. Friedman they became respect- 
ed for their use of statistical methods. 4 Much of 
their time and expertise was devoted to creating 
codes and ciphers for the army. But they devoted 
an increasing amount of effort to operational 

(U) Although separate from OP-20-G, the SIS 
had a gentlemen's agreement about cryptanalytic 
turf. Friedman's group agreed to focus on enemy 
army systems but to share a rather ill-defined 
zone of diplomatic and clandestine traffic with 
"G." The Coast Guard's cryptanalytic office, led by 
Friedman's wife, and the FBI's codebreaking 


Page 127 


group shared in tapping the diplomatic and clan- 
destine traffic in the Americas. Like OP-20-G's 
crew during the 1930s, the SIS's men were direct- 
ed to concentrate on Japan's secret systems but 
not given the resources to fulfill the charge. 

(U) The difficulty of intercepting enough mil- 
itary messages extended to the SIS's attack on 
Japan's army systems. Unlike the use of high- 
powered radio by the navies, the armies and air 
forces of the world used low-power systems and 
sent relatively few messages that could be inter- 
cepted from a great distance. Even after the SIS 
constructed listening posts in the Pacific and the 
Canal Zone, 5 it could not acquire military mes- 
sages in enough "depth" for code or cipher break- 
ing. 6 

(U) As a result, Friedman's talented men and 
women spent much of their time during the 1930s 
on diplomatic communications. After months of 
intense work, in 1940 they laid the foundation for 
America's Magic by successfully attacking 
Japan's new Purple enciphering machine system. 
It carried Japan's most important diplomatic 
messages to and from the world capitals. 
Although Friedman's group received help from 
the navy in attacking Purple, Magic was seen as 
an SIS triumph by the nation's leadership. 

(U) Friedman's group had employed modern 
as well as traditional cryptanalytic techniques 
against Purple. A few years after OP-20-G began 
to use tabulating machines, the SIS established its 
first automation foothold. 7 Although it did not 
begin an OP-20-G-like Rapid Machine project 
before the war, the SIS hired a newly minted MIT 
electrical engineer at a critical stage in the 
Japanese diplomatic problem. That graduate of 
MITs electrical engineering department, Leo 
Rosen, helped break into the Japanese diplomat- 
ic machine and constructed its first analog. 8 

(U) Although William F. Friedman's group 
had ideas for teletype-tape comparators, iso- 
morph machines, and relay attachments for tabu- 

lators, it did not have the resources to turn them 
into hardware. It did not go beyond building 
direct analogs of enemy machines. 

(U) The Search for Another American Ultra 

(U) When war broke out, the SIS had little 
cryptanalytic capability, few intercepts, and little 
machinery. It had no Enigma proficiency, it was 
unable to read the major Japanese or German 
military codes, and it had few messages or 
machines with which to analyze them. In fact, it 
appeared that it would be some time before the 
SIS would have much to work on. 

(U) It did get one assignment but through 
default. "G" was overworked because of its efforts 
against critical German and Japanese naval sys- 
tems. Out of necessity, it turned all of the 
Japanese diplomatic problem over to the SIS. 9 
That Purple diplomatic challenge took much of 
the army's attention in the first year of the war, 
although the system had already been solved. 
Purple had become a relatively easy system to 
exploit. It needed a few new electromechanical 
analogs, but it demanded little else. Even Britain 
cooperated. It sent intercepts and cryptanalytic 
advice to Friedman's Japanese experts. The 
attention was well rewarded. Much was learned 
about Germany, as well as Japan, from the radio 
and cable messages to and from Japan's 
embassies in Axis and neutral nations. 

(U) The German problems were very differ- 
ent. Its diplomatic systems proved difficult to 
enter, and its military codes and ciphers resisted 
attack. As important, the British, who held many 
secrets to entering Enigma and other German 
ciphers, did not wish to grant the SIS power over 
Germany's army or air force systems. 

(U) The SIS badly needed Britain's help. It 
began World War II with as little, perhaps less, 
potential to enter German systems as OP-20-G. 
When the SIS finally decided to establish an 
Enigma program and demanded to become a 

Page 128 

TOP GCGrcrn/COM IH TOnCL TO USA, aus, cam , qbb AND I4ZI mi 


partner in Ultra, it found that it had little to nego- 
tiate with. Its main bargaining chips, Purple and 
Magic, had been given away in early 1941. 10 

(U) The SIS had a much more difficult time 
than OP-20-G in gaining GC&CS's trust. 
Throughout the war the SIS men felt they had to 
fight much harder than the navy for British con- 
cessions on Ultra." They worried that the British 
promises of full cooperation that had been made 
as early as the autumn of 1940 might never be 
kept. In a way, their fears were correct. GC&CS 
never granted the United States Army's cryptana- 
lysts as much independence as it did the navy's 

(U) And while playing atug-of-war with 
Britain to gain knowledge of the German ciphers, 
the SIS was tormented by the Japanese army 
code problem. It was not until the spring of 1943 
that the SIS centers in Washington and Australia 
were able to tap a major army system. 12 Perhaps 
it was the need to devote its energies to the 
Japanese codes, and a belief that traditional 
methods were the only alternative for such prob- 
lems, that led the SIS to be much later than the 
navy in establishing a formal group to develop 
rapid machines for statistical and mathematical 

(U) Its delayed start led the SIS to rely on the 
navy to supply most of its initial RAM equipment. 
But it then launched perhaps an overly ambitious 
attempt to create a very advanced RAM, one that, 
it was hoped, would leap-frog the navy's technol- 

(U) A Great Electronic Adventure, the Freak 

(TO//0I) Like the ex-MIT engineers at OP-20- 
G, some of the SIS's technicians had great faith in 
advanced electronics. Their first dream at the 
outbreak of the war was for a new type of machine 
to perform one of the most tedious but important 
general cryptanalytic functions, frequency count- 
ing. Their goal was to create a relatively small and 

super-fast machine to count and record all simple 
and digraphic frequencies. The machine was to 
do that in one pass through a message. 
Tabulators, because they had so few counters, 
demanded much sorting and many card runs to 
complete a full count. Many of the standard tabu- 
lator frequency-counting procedures used in the 
attacks against Purple, Hagelin and, later, some 
teletype systems took sixty to ninety hours. 13 

(U) To go beyond the "tabs" for such complex 
counting was a challenge. For one thing, it called 
for the creation of a new type of memory. Readily 
available technological options, such as using 
industrial counters to store results, meant accept- 
ing slow processing and a machine that would be 
the size of a room. 14 

(TS//SI) In mid-1942 Leo Rosen decided to 
take up that challenge. Recently put in charge of a 
group of engineers, he decided to establish an SIS 
RAM program. He thought that it should begin 
with fundamental contributions. He was deter- 
mined to develop a large high-speed electronic 
memory. In addition, he told his men to create 
electronic circuits that could perform analytic 

(U) Leo Rosen 


Page 129 


functions, such as a sigmage test, that were an 
integral part of all the frequency-based crypto- 
attacks. 15 Together, he hoped they would provide 
the basis for the high-speed counting machine, 
Freak, a device that OP-20-G seemed unwilling to 

(TS// 8 I) Unfortunately, Rosen picked a much 
too ambitious goal for the army's first RAM 
adventure. It took a year and a half before Freak I 
emerged from the SIS workshop; then, it proved 
too delicate for operational work. 

(U) Freak 

(T0//0I ) Freak's design and components were 
major advances in the technology of calculation. 
Rosen's group had decided to use more than 
7,000 condensers for the machine's mass "mem- 
ory." One thousand twenty-four sets of seven con- 
densers each were the "counters" in Freak. ' 6 By 
using the binary counting system, every set could 
hold a count of up to ninety-nine. The enormous 
number of condensers accounted for much of the 

size of Freak I. The machine was nine feet high by 
eight feet long. 1? 

(TS//6I) There were great hopes that Freak 
would speed all types of counting and analysis. As 
the data were read in from two tape readers and 
processed through a relay system, the appropri- 
ate counters were incremented. Then the 
advanced electronic digital circuits calculated 
running frequencies and the critical SUM 
(N(Ni)/2). The circuits also scanned the counters 
and controlled an electromatic typewriter which 
printed the results. 

GSyyef) The use of the bina- 
ry system, the memory technol- 
ogy, and the digital calculations 
were advanced for the time. 18 
Freak I, unfortunately, did not 
have along or useful life. The 
counters proved unreliable, and 
the electronic circuitry was 
troublesome. The machine was 
so uncooperative that it was dis- 
mantled in mid-1944, just six 
months after its birth. 19 

£¥9} The defeat on the first 
Freak tempered the SIS's faith 
in electronics, but its engineers 
decided to try again. They con- 
structed a second version that 
began twenty operations in 
spring 1945. It proved more reli- 
able than its predecessor. "° 

(U) Tabulators and Traffic: A Data 
Processing War 

(U) Despite the affection for electronics, the 
SIS placed its faith in older technologies during 
the first critical months of the war. The SIS made 
an agreement with IBM and soon had scores of 
tabulating machines. Many IBM engineers were 

Page 130 

Tnr MCBBaco T "" CI Tniirfl fl " r r "" rnn > M n u 7 i , m 


sent to Washington to make significant modifica- 
tions to the tabs and sorters, and IBM's factories 
were kept busy producing special devices. By the 
end of war, the SIS had close to 400 IBM 
machines using a million IBM cards a day. 

( T3//CI j- The workforce for the machines 
grew from fewer than 100 at the end of 1942 to 
600 a year later. By the end of the war the SIS's 
tab rooms had close to 1,200 workers. 21 

E B//SI) 'As the SIS waited for intercepts from 
the Japanese military systems and hoped for 
information from the British on the Enigma, they 
did their best to produce intelligence from the few 
sources besides Purple that were available to the 
agency. 22 

CIS//0F) One ofthose sources was the inter- 
cepts of diplomatic messages sent on Germany's 
GEC system. The army's radio men had been col- 
lecting them on their own for some time, as well 
as receiving information on them from the 
British. But collecting was easier than solving the 
system. The GEC codes were tough. The Germans 
used code words doubly enciphered with addi- 

(T9//9"I ) The tabulator group at SIS began an 
attack on the system using labor-intensive tech- 
niques similar to those the navy had developed to 
breach the Japanese navy's additive systems. The 
going was difficult, however. The usual attacks 
did not seem to work. The Germans had a very 
clever keyword system for specifying the addi- 
tives that proved difficult for the Americans to 
penetrate during their first year of IBM attacks. 
Fortunately, the British had acquired some pages 
of additives from a French agent and decided to 
pass them to the Americans in early 1942. 

(TO/SI) With the hints about the system, the 
Americans launched their first new tabulator 
attack of the war. Their work on GEC led them to 
develop machine methods, such as the search for 
double repeats, that were transferred to the 

Japanese military problems once a flow of inter- 
cepts began. 23 

gS// 8 J j But the SIS had to wait quite some 
time before the army could supply enough 
Japanese material. Then the SIS cryptanalysts 
found that Japan's military had, perhaps unwit- 
tingly, been wiser than its diplomats. Japan's 
diplomats had made a mistake by basing their 
secret communications system on a machine. By 
turning to the latest technology, they had made 
their ciphers more vulnerable than if they had 
used, for example, crude one-time pads. 

£ K)/0fl In contrast, the Japanese military had 
decided to stay with older methods. In doing so, 
they frustrated the British and American code- 
breakers and forced them to turn to very "data 
heavy" methods. The Japanese army's code-with- 
additive systems were vulnerable to capture, but 
neither the Americans nor the British acquired 
any significant amounts ofmaterial during the 
first years of the war. 

QCS//8I) A cryptanalytic attack without cap- 
tures, or quite evident "busts," demanded enor- 
mous numbers of intercepts, analysts, and 
machines. It meant that the SIS had to engage in 
a frustrating data processing war. 

(TS//SI) Despite the allocation of massive 
amounts of resources to the problem, the 
Japanese army's systems resisted longer than its 
navy's. The difficulty of intercepting its messages, 
its use of complex additive systems, and its clever 
ways of hiding the information contained in the 
message preambles led to a near cryptanalytic 
blackout during 1942 and 1943. 

CDS// S I) The inability of the Allies to read the 
major army systems through cryptanalysis led the 
SIS to rely upon traffic ^is; ia conse- 
quence, IBM tabulators and methods quite like 
those in business data processing became essen- 
tial to its operations. 24 Throughout the war hun- 
dreds of machines and people were kept busy 


Page 131 

T I 1 P .W.M i m-n Turin T n m^. , , ^ |rr ^ nnn A , |n ^^ 

sorting, counting, and listing frequencies of com- 
munications among units. Even when some of the 
major army systems had been penetrated, 20 per- 
cent of the total machine hours in SIS were devot- 
ed just to the analysis of the message headings. 

CTO//0I) Keypunching and the physical main- 
tenance of card files for the traffic analysis pro- 
cessing were demanding by themselves. 
Recording and analyzing 300,000 messages a 
month for traffic analysis was not uncommon. 
Huge decks of cards had to be carefully loaded 
into the tabs and sorters for the first of the many 
steps in each analysis routine, then reloaded sev- 
eral times to complete a process. The pressures all 
that created were so great that the SIS Machine 
Branch had to endure a critical personnel prob- 
lem: after several weeks oftraining, machine 
operators quit. The young civilian women were 
apologetic, but insisted they be allowed to leave. 
The night shifts were especially difficult to staff, 
and it was only the arrival ofWACS, who were 
allowed to live on base and who were unable to 
resign, that allowed the machine room to contin- 
ue its twenty-four-hour work day. 25 

(V) Making the Tabs More Powerful 

£58-) The IBM tabulators remained the foun- 
dation of the SIS's operation and for more than 
traffic analysis. The continued dependency on the 
tabs was reflected in the intense efforts the SIS 
made to increase their power. By the end of the 
war, the Arlington Hall engineers had helped to 
develop an impressive array of specialized IBM 
equipment as well as a whole series of complex 
relay attachments. ** 

ff S//3 ' f) ' Modified "tabs" were the technology 
for the SIS's work on Japanese code problems. 
Some twenty-three different relay attachments 
were used to attack Japanese army systems. In 
many instances, the attachments were so power- 
ful that the "tabs" were relegated to being mere 
input-output devices serving the relay cabinets. 

( TS//SI) The complex relay circuits placed on 
the tabs automatically stripped additives, applied 
possible encryption squares, and even searched 
to see if the results of additive removal had led to 
the appearance of high-frequency code groups. 
The more complex attachments received names, 
such as Brute Force, Camel, JMA, the Selective 
Square, or the Limited Selector. 

(T9//3I) T he modifications that automatical- 
ly decoded upwards of 2,500 messages a day were 
of special pride to the SIS's machine branch. They 
allowed the timely exploitation of the captures of 
cipher text and key that began to flow into 
Arlington Hall during 1944. 2? 

fFS//SI) One of the most impressive of the 
SIS's tabulator modifications was the Slide Run 
machine. Tts origins illustrate why most of 
America's codebreaking history is so unlike the 
thrilling story of MAGIC, when a brilliant insight 
supposedly led to a near instant victory over a 
major communications system 


(U) Slides, Runs, and Endless Decks of Cards 

(T0//0I) The mature Slide Run machine of 
mid-1944 was a combination of a 405 Tabulator 
and a huge chest-high, multipanel cabinet full of 
advanced relay circuits, telephone crossbar 
relays, counters, and plugboards. But the Slide 
Run did not begin as an elegant example of IBM's 
best work. The. first two of the devices were hur- 
riedly built by the "F" Branch in late 1943 as an 
emergency response to the requests of the crypt- 
analysts. The "cryppies" thought they might have 
found the techniques and some of the additive 
keys needed to attack the Japanese army systems. 

CTO// 31) Although hastily built to exploit a 
particular opportunity, the machines proved so 
valuable that a development and production con- 
tract was signed with IBM. The new Slide Runs 
were to be used on a variety of problems. By early 
1944, IBM was constructing six more Slide Run 

Page 132 



machines, each more sophisticated than its pred- 
ecessor. ^ 

(TS//SI) The Slide Run machines were badly 
needed because of the continuing difficulties with 
Japanese army systems. The Japanese army 
problem was very difficult and all attacks were 
extremely labor intensive. Hundreds of cryptana- 
lysts at SIS had been working since the beginning 
of the war to discover the numeric additives and 
the codegroup meanings. The tabulators had 
been called upon to process files of as many as 
3,000,000 cards. 

(T8//9I) Copperhead-like brute force search- 
es, keyword searches, and repetitive additive 
stripping and testing kept machine-room double 
shifts busy for months. Despite all the effort, 
there was little more than frustration until April 
1943 when the Japanese army's indicator system 
was broken. That allowed the identification ofthe 
enciphering squares that were used in the indica- 
tors. With that breakthrough, it was possible to 
identify messages that were enciphered using the 
same additive pages. With knowledge of the 
pages, an attempt could be made to place mes- 
sages on overlaps by tabulator-based, brute-force 

(TS//SI) At first the search for the repeat of 
the same cipher groups at the same intervals in 
two messages (brute force) had to be done 
through the tedious repetition of card duplication 
and endless sorting.The most efficient attacks still 
demanded files of almost 200,000 IBM cards. All 
that processing was tolerated just to try to find 
messages that had a probability of being enci- 
phered with the same set of additives. The job 
became too much, even for the SIS's hundreds of 
tabulators and sorters. Away to automate the. 
process had to be found 

QXS//&B In response to the need, a relay 
attachment was built for the brute-force search; 
but the Japanese army attack continued to 
demand round after round of card punching, 

reproduction, collating, sorting, addition, sub- 
traction, multiplication, and printing. For exam- 
ple, one procedure developed in the summer of 
1943 hoped to recover additives by using four 
copies of a file of 430,000 cards. The files were 
run three shifts a day until the end of the year. 

(TS//SI) By fall 1943 the many tabulator 
attacks and an increasing number of captures of 
Japanese material allowed the accumulation of a 
file of additives. The next logical steps were to try 
to locate the messages that used particular addi- 
tives and to strip the additives to reveal clear 
numeric codes. From there, the cryptanalyst 
could recover more code meanings and decipher 
more messages. 

(-E9//STJ The automation of the process came 
gradually. At first, the "slide run" procedure was 
just a new way to use the existing tabs. It was 
developed to replace the old hand methods of 
applying a known additive to a length of code, 
then testing to see if a sensible result emerged. If 
an unlikely group appeared after the additive had 
been removed, the additive was tried against the 
next offset of the text. If a juxtaposition of addi- 
tive and cipher yielded a likeiy result, it was test- 
ed against a file of frequently used code groups. 
When a match occurred, the cryptanalysts con- 
cluded that it was probable that the additive they 
were trying might have been used to encipher the 
current message. 

(TS^tfI> The "slide" was a common sense but 
powerful technique. But even when it was imple- 
mented in IBM methods, it stretched human as 
well as machine resources. When regular tabula- 
tor equipment was used, the routine was quite 
complex. A series of five likely keys (additives) 
was punched on cards so that they could be test- 
ed against all possible positions in thousands of 
messages. When the key produced, in two of the 
five tested positions, code groups found among 
the already known most frequent 250, the tabula- 
tor printed the message number and the five code 


Page 133 


CID//QI ) Slide Run 

(T3//3D W hat came to be called the Slide 
Run machine was developed to reduce the size 
and number of required card files and to speed 
the "slide" testing process. The first step towards 
the eventually very sophisticated device was the 
invention of a code-recognition component. To 
further automate the process and to reduce the 
number of cards that had to be handled, aeon- 
version unit was constructed. Its relay circuits 
stripped the additives to produce the code sent to 
the recognition unit. 

if it completed any of the code circuits. If it did, 
the machine recorded a "hit." 30 

(TS//S*) More and more "intelligence" was 
built into the Slide Run machines. New models 
appeared which included sensitive and labor-sav- 
ing statistical threshold tests. They prevented the 
printing of unprofitable reports. The first of the 
postwar versions went further. It used a log- 
weighting method to reduce the number of 
"prints." 31 

(TS//SI) In the Slide Run machine, banks of 
relays were wired to hold as many as 250 code 
groups; later versions held as many as 1,500. The 
recognition unit tested each stripped group com- 
ing from the tabulator's arithmetic section to see 

(TS//SI) The Slide Runs were of great value to 
SIS, but they were not exceptionally fast nor eas- 
ily "programmed." The best of the World War II 
versions read cards at a rate of 150 a minute, and 
it took from three to five days of work to set the 

Page 134 



codes in the recognition units. Then it took hours 
to run the cards for a series of messages. 32 

(TS//SI ) The impossibility of increasing the 
rate of card-sensing much beyond what had been 
accomplished by 1944 is what led the SIS, in June 
1945, to make a request for a RAM 70mm film 
version of the three "tab" machines that had 
proven so useful against the Japanese systems: 
the Slide Run, the Isomorph, and the Brute Force 
machines. Unfortunately for the men within Leo 
Rosen's section who, as we will see, became the 
torchbearers for Bush's ideas, the army decided 
the Eastman designs should not be funded. 33 

(U) The Other Bombe Program 

(U) Well before any progress had been made 
on the Japanese military problems, the SIS decid- 
ed that it had to gain a share of the European 
Ultra. And like the navy, it wanted control over its 
own intelligence resources. But it was ill prepared 
in terms of skills, equipment, or political power 

(TS// S I) The SIS did not begin its Enigma 
battle technologically prepared, nor did it have 
any plans for advanced anti-Enigma machines. At 
the beginning of the war, the SIS's men were told 
of OP-20-G's RAM contracts with Eastman and 
Gray. As a result, Rosen and Friedman became 
interested in the possibilities of microfilm-based 
machines, and they agreed to at least examine 
their possibilities. Soon, they heard a few things 
about the navy's Bombe ambitions. 34 

ffiG/ /3Y T But during the first months of 1942 
the SIS focused on expanding its tabulator sec- 
tion; it was not until late in the year that it decid- 
ed to create a machine research group that had 
the manpower needed to examine, let alone cre- 
ate, new technologies. The MIT graduate, Leo 
Rosen, was placed at the head of a small team that 
began its work in the basement of an old house at 
Arlington Hall Station. 35 One of his first actions 
was to advise his superiors that the SIS should 
join in the RAM program. He convinced the army 

to purchase almost $200,000 worth of copies of 
the OP-20-G-sponsored machines from Eastman 
and Gray. 36 The army gave the Bureau of Ships 
the funds needed to purchase machines similar to 
Tessie, the IC machines, and, later, a Gray-NCR 
Comparator. Letterwriters were also supplied by 
the navy. As within "G," they became an essential 
part of SIS's data processing services. 

(U) Another Step Back 

d SS//8f ) Rosen's major assignment, however, 
was to produce a machine to give SIS the kind of 
power the army thought OP-20-G was gaining 
over GC&CS through its emerging Bombe pro- 
gram. By summer 1942 he gained approval for an 
SIS Bombe program. He began to assemble a 
force of enlisted engineers and technicians but 
soon realized that the army would be unable to 
build or even design a Bombe by itself. 38 His staff 
remained too small through 1942, and he had to 
confine its work to preliminary investigations and 
minimal construction projects. 

(U) One of the first of its preliminary studies 
was on the possibility of a new type of Bombe for 
the army and air force Enigma problems; that 
turned out to be a major task. Soon Rosen 
thought enough had been learned to allow a com- 
mitment. In October 1942 the SIS decided it had 
to have its own version of a Bombe, and it was to 
be acquired independently of Britain and OP-20- 
G. 39 

■ (0) R osen's "F" team explored an electronic 
Enigma while Friedman made the rounds of the 
scientists associated with the NDRC's fire control 
computers. 40 Rosen's electronic option would be 
put aside for the same reasons OP-20-G had 
dropped electronics during the summer. But the 
SIS's hope for a tube-based solution lasted a bit 
longer. 41 

(T S // S I) Belief in the potential of electronics 
led to Rosen's team hiring experts from the tele- 
phone company. They worked on an electronic 


Page 135 


AS//3I ) Madame X 

Bombe until December 1943. Then frustration 
with the disappointing results led to an end of the 
contract. However, SIS's faith bounced back, and 
it began the hunt for another high-speed rotor 
after its "E" crisis had passed. 42 

(U) While Rosen and the Western Electric 
experts explored electronics, one of the alterna- 
tives recommended by the NDRC's researchers 
was approved. What became known as Madame 
X followed the general logic of the Turing attack 
on the Enigma, but it was significantly different 
from the British and the OP-20-G Bombes. The 
machine first appeared as a breadboard demon- 
stration unit in early 1943, passed its first tests 
that summer, and was available as an operational 
model in October 1943. 43 

(8//SI) Madame X(also called "003") was 
huge. It was so large because the SIS had decided 
to be more elegant and innovative than the navy. 
It wanted one grand Enigma-fighting machine, 
300 uncoordinated ones. Although it did not 
meet all its original goals, the "003" was an 
impressive machine. 44 It contained 144 Enigma 
scrambler units, as compared to the sixteen in the 
standard OP-20-G Bombe. "003's" banks could 
be divided into different size groups so that as 
many as twelve useful menus could be run at one 

time. 45 The army cryptanalysts knew that the 
longer a crib and its chains, the fewer the false 
hits. A large number of units linked through a 
flexible central control system would allow sever- 
al shorter cribs to be run simultaneously. 46 

( TO//OI) The "003" was designed for more 
than simultaneous runs, however. It was intend- 
ed to be easier to use than the navy Bombe and to 
have a much faster setup time. Some of those 
goals were achieved. When the "003" was com- 
pleted, the army's technicians sat in a "turret" 
room and set plugs and switches on small control 
boards rather than having to place dozens of com- 
mutators on the machine as the navy's operators 
had to. 

(T3//3I) The "turret" system was quite inno- 
vative. There were a dozen of the switching sta- 
tions in the room. Each "turret" was a small ver- 
sion of a telephone switchboard with an addition, 
a set of push buttons. Each board could control its 
own part of the "003" if simultaneous runs were 
desired. Such simultaneous use seems to have 
been the norm once the SIS learned that well- 
selected short menus could be powerful. Strong 
menus sent from England allowed the use of only 
a few of the "E" units (perhaps fourteen) per test. 
Typically, some ten problems were run atone 

Page 136 



time. On its best days, "003" completed 1,200 of 
the short three-wheel runs." 47 

ffi3//SJQ -The switchboards were used to select 
which subset of the frames were to be active dur- 
ing a run. The push buttons were truly a unique 
and potentially valuable feature. Through them 
wheel orders could be changed in one-half a sec- 
ond. That allowed wheel orders to be tested in 
rapid succession. Unfortunately, the other parts 
of the setup, including the menu, had to be done 
by hand on the individual Enigma frames. That 
led to the setup time for the "003" being much 
more than hoped for - some twenty minutes for a 
test when more than wheel orders had to be 
changed. At peak efficiency, a crew could place 
and run twelve new menus a day.* 8 

(U) Avery innovative and important feature 
of the "003" was its ability to automatically con- 
trol the stepping motion of its "wheel" analog. 
The relay circuits allowed the machine to use 
"non-metric" motion. 49 

(U) More to It Than the Madame 

f&) Like the navy's Bombe, Madame X could 
not work alone. Before it could be efficiently used, 
the SIS had to have cribs and a list of what wheel 
orders would not have to be tested. Then, after a 
"hit" was found, the SIS had to do as much or 
more hand-testing than the navy. Several 
machines to speed the hand work were built. 
They performed the same type of functions as the 
navy's M8 and M9. 

( TS/filj The army needed a special aid to help 
"003" because of what might be called a design 
oversight in late 1942. The design of Madame X 
had begun before the SIS knew all about the 
British Enigma attacks and before it had enough 
experience to realize how necessary it was to have 
machines that eliminated all but a very, very few 
possible keys and settings. Thus, the original 
Madame X did not have a full "diagonal board" 
test built into it. 

( TO// SI ) SIS's men thought their bombe 
would be useful even though its list of "stops" 
would not be filtered by a thorough test for steck- 
er inconsistencies as was found on the British 
Jumbo Bombe. As a result, they thought that a 
celluloid grille would be sufficient to search for 

(T0//SF ) That was incorrect. That hand "diag- 
onal board" test proved so time-consuming that 
the first of the SIS's versions of grenades was con- 
structed. 50 But as more was learned about the 
navy's and GC&CS's machines, Rosen's team 
decided to build an attachment that would auto- 
mate the entire consistency-checking process. 51 

( TS//SI) The American Machine Gun had its 
first tests in September 1943, a month before the 
second half of "003" was completed. Built of the 
same technology as its host, the Machine Gun 
searched for stecker inconsistencies and sup- 
pressed all the "stops" that were logically impos- 
sible given the nature of the Enigma plugboard. 
The "Gun" speeded SIS's work but caused some 
discomfort for those assigned to "003's" rooms. It 
was a very noisy device, as was what had inspired 
it, Britain's older grenade. 52 

tSkSome other special devices were attached 
to the original "003." like many of the navy's 
grenades, the army's add-ons took advantage, 
whenever possible, of German procedural errors. 
But some of the attachments and alterations were 
designed to apply new general cryptanalytic 
knowledge to speed "003" processing. 53 

ggi//SI) Many of the ideas for the new attach- 
ments came from the British, who were creating 
similar devices for their Bombes. Their CSKO 
switch, for example, had been very helpful. Once 
on "003," it helped exploit a German air force 
procedural rule forbidding linking any checker- 
board letter to its neighbor. For example, B could 
not be checkered to A orC. The Consecutive 
Stecker Knockout circuits checked for such illegal 


Page 137 


connections and prevented a "stop" from being 
indicated. 54 

(TS//SI) Madame X was given another help- 
ful attachment, the Double Input. It was an appli- 
cation of more sophisticated knowledge of the 
probability aspects of cribbing. This allowed two 
relatively weak menus to be run simultaneously 
and to approach the power of the usual sixteen- 
letter menu with "closure" on the crib-plain com- 
binations. 55 

0TO//OI) Alterations to "003" allowed it to 
drag cribs so that it could am Swiss and Spanish 
Enigma problems. And the Clambake attach- 
ments were for short runs for "grenade cribs." 
The "003" was allowed to perform more flexible 
tests by using the Oyster Schuker attachment. In 
addition, methods to exploit Cillys were devel- 
oped, and "003" was, at times, rewired to handle 
special Enigma reflector runs. 56 

4&)r There were several attempts to make 
Madame X more powerful; some were very suc- 
cessful. A difficult goal was to expand the powers 
of "003" by lessening its dependency on cribs. 
The first appearance of the idea for a probability- 
based way to find wheel settings, as well as to deal 
with garbled cipher, emerged before "003" was 
completed. 57 Taking similar British methods into 
consideration, 58 the SIS cryptanalysts and engi- 
neers decided to use thirty-six of "003's" own "M" 
frames to test for the frequency of appearance of 
the sixteen most unlikely letters within a fifty-two 
letter test decipherment. Just the cipher text was 
entered. If fewer than nineteen of the fifty-two 
letters that resulted from a deciphering at a par- 
ticular starting point were on the high frequency 
list, a "stop" was printed. 59 

(T9//3I) Unfortunately, such use of "003's" 
regular frames led to the machine being monopo- 
lized by special tests. One of the worst of them 
was "dudbusting." Duds were messages that 
should have been readable given knowledge of 
their keys, but were not. They were usually the 

result of operator errors such as the use of an 
incorrect key on a system. 

ff0//3f) Because dudbusting was one of the 
more important tasks assigned to the "003," and 
because the job took so long, a decision was made 
to create a new dudbuster with its own "frame," 
and, perhaps, a bank of electronic counters. 60 The 
electronics proved a bit too much for an emer- 
gency situation, but a more efficient and quite 
impressive electromechanical (relay) Arlington 
Dudbuster became an essential part of the SIS's 
machine rooms. The Dudbuster worked on the 
principle of recognizing plain language through a 
simple frequency test. Based on the characteris- 
tics of the German language, including the 
absence of "X," each plain letter was assigned a 
weight. If the summed weights equaled a thresh- 
old value, text was printed, then examined, to 
see if true German was a result of the wheel 


(T S // S I) There was also a film RAM version 
of the SIS statistical Dudbuster. Its birth led to 
some friction between SIS and OP-20-G. 
Independently of OP-20-G, the SIS cryptanalysts 
conceived of their own film Hypo. When they 
informed OP-20G of their great discovery in mid- 
1943* "G's" men became quite embarrassed. They 
were forced to admit that they had thought of the 
Hypo method earlier, had a machine in develop- 
ment, but had not informed either the SIS or 

(TS//SI) Despite some help from the navy, it 
took a long time for the SIS to turn their Hypo 
dudbuster idea into hardware. It was not until 
late 1944 that a special camera was linked to an 
"M" frame to generate the images of the distribu- 
tion of high frequency letters from each Enigma 
setting. Those master films were then run, as in 
Hypo, against cipher text to find the point of 
greatest coincidence. 62 

( TS//SI) The film dudbuster and Hypo were 
often called "Grenades." Their success led the SIS 

Page 138 



to explore the possibilities for more ambitious 
film attachments and supplements for "003." 
Unfortunately, some of the most ambitious 
Grenade ideas could not be implemented. The 
high-speed Azalea and Bachelor attachments 
were never completed, and the plans for a super- 
speed film cribdragger probably did not turn into 
a project. 63 It is certain, however, that neither the 
army nor navy ever built a Grenade that matched 
the grand achievement of the British in late 1943, 
its electromechanical Fillibuster. It tested four 
cribs on eighty messages simultaneously. 6 * 

(TS//SI) The alterations to Madame Xdid 
make it more efficient, but the American army's 
Bombe continued to make many demands on the 
SIS. Its manpower requirements were not as 
great as the navy's Bombe, but it was a labor- 
intensive machine. It needed more than twenty 
operating personnel and forty maintenance men. 

■$¥&] Despite the large maintenance crew, the 
machine had a bit of a tendency to lose its con- 
centration. During its first few months, all runs 
were duplicated on a second control board. And 
well into 1944 it could not be coaxed into running 
at the originally hoped-for speed. It had been 
designed to run, if desired, at sixty pulses a sec- 
ond for short periods, but its typical operating 
speed was half that, thirty pulses per second. 65 

(U) Thus, despite its great flexibility, "003" 
was not a perfect solution. Its successes certainly 
did not change the navy's mind about Bombe 
architecture. When OP-20-G decided to build a 
second set of fifty machines, it found Desch's 
commutator design much faster and more effi- 
cient than a relay-based machine. 66 

(U) "003" could not be switched to a four- 
wheel mode, and twenty-six separate runs had to 
be done to test a four-wheel message. As objec- 
tionable to the navy's engineers was "003's" slow 
operating speed. The automatic control system in 
Madame X did not fully compensate for its long 
run times. 

(U) The "003" had another feature that did 
not prove as powerful as hoped. The army 
machine had the circuits necessary for the "loca- 
tor" task done separately by the navy's Hypo 
machine, but the navy's engineer-cryptanalysts 
did not find "003's" automatic locator that attrac- 

(TS) Near the end of the war, Joseph Wenger 
requested another comparison of the army and 
navy "E" machines. Wenger decided that the OP- 
20-G Bombe complex was, on average, fifty times 
more productive than the army's machine in 
Arlington. 67 

(T0//0I) The main reason why only one "003" 
was built was the SIS's inability to convince the 
British to yield the messages and techniques 
needed to keep the machine busy. 

(U) A Machine Looking for Work 

(TS//6I) Madame X's construction began just 
as GC&CS and the American army reached a low 
point in their relationship. Britain's refusal to 
inform the SIS about anti-Enigma methods led to 
an impasse. The situation became quite tense. 
The Americans, for example, refused to let Alan 
Turing see their new voice scrambler system until 
Britain yielded her secrets. For a time, SIS even 
withheld what it was discovering about Japanese 

( TS//SI) Friedman's group certainly must 
have resented Britain's request that it build a 
super-Bombe for GC&CS while she would not 
share her cryptanalytic secrets. There must have 
been complaints about being asked to construct 
advanced machines to attack the Fish system's 
Geheimschrieber yet not being allowed to have 
the desired intercepts and cryptanalytic informa- 
tion on Enigma. 


(TS// S I) The tensions were reduced in May 
1943 after a series of conferences, but there was 
no immediate flow of cryptanalytic secrets to 


Page 139 


America, nor was there enough high-priority 
work assigned to Madame X. The result was that 
it seemed to be an expensive machine without a 
purpose. Then, when it was finally put to work, it 
served as a secondary aid to GC&CS. 

ffW The SIS had not expected such a 
minor role for their great computer. Its best 
cryptanalysts had spent much of 1942 and 1943 
trying to develop their own "E" attacks, using up 
1,000 hours of "003" time. But after all that effort 
they had to admit defeat and had to yield to 
England's monopoly of methods and its control of 
intercepts. The home grown 'Yellow" project for 
which "003" was built became asubunit of 
Bletchley and, to make matters worse, most of the 
SIS's involvement with "E" came through the 
team of men it sent to England in late 1943. 
Americans were as busy running British Bombes 
in a small town near Bletchley as they were run- 
ning Madame X in America. Perhaps as depress- 
ing for Rosen and his group, England was proba- 
bly sending as many German air force jobs to OP- 
20-G as it was sending to Arlington Hall. 69 As a 
result, Madame X stood idle much of the time. 
For many days it had to be assigned only to 
"research problems." 

BS^- Such disappointments with the SIS 
Bombe program were perhaps what forced 
Friedman to write a very defensive report about 
Madame Xin early 1944. He tried his best to 
show that Madame X was superior to all the other 
anti- "E" devices. He hammered at every weak- 
ness of the commutator Bombes, British as well 
as American. They were susceptible to mechani- 
cal stress, they were not good at solving "duds," 
and they took much too long to record the "hits," 
he said. He continued by citing "003's" need for 
fewer maintenance personnel and by stating that 
the British Bombes took twice as long as "003" to 
run an entire problem. 

£ES}*ut he could not hide the deficiencies of 
Madame X. He admitted that at the very least, the 
"003" was five times as expensive as the British 

Bombes and that it was really slower than the 
American navy's machines. He tried to conclude 
his report on an upbeat note, but his final state- 
ment helps explain why Madame Xwas torn 
down at the conclusion of the war: 70 "It is thought 
fair to state that, for purely operational purposes, 
the rotary type of bombe is the most efficient but 
for research and development of new solution 
methods, the relay type, because of its greater 
flexibility, is far superior." Madame Xdid win 
some victories, however. As an engineering proj- 
ect and as an example of America's mass produc- 
tion capabilities, it captured the respect of the 
British engineers. It also served as a backup sys- 
tem for GC&CS. It was kept busy from 1944 to the 
end of the war running noncritical or stubborn 
jobs sent to it by the British. 

( TS//SI) "003" cost only one-fifth of the year- 
ly construction budget for a system that also 
found itself with much less than the expected 
amount of work. The huge voice encryption sys- 
tem the Signal Corps and the "F" section were 
building, the Sigsaly, had a $5,000,000 budget in 



(U)Asa result of "003's" demonstration of 
the SIS's engineering abilities, Britain asked the 
agency to help solve more complex engineering 
problems and gladly accepted some American 
machines within GC&CS. 72 By the end of the war, 
the SIS was busy creating very advanced 
machines and, like OP-20-G, was hoping for a 
permanent in-house computer program. 73 

(U) More Emergencies and More 

(U) The engineers at"G" and the SIS may 
have wanted to launch a far-reaching electronic 
development program at the end of 1943, but the 
Allies faced too many cryptological crises to allow 
the pursuit of any grand ambitions. Emergencies 
continued to drive the efforts at NCR and 
Nebraska Avenue, as well as at Arlington Hall. 
Some of the crises dictated the use of very 

Page 140 


EO 3. 

3(h) (2) 


advanced technology, but others were handled by 
relying upon traditional components and archi- 

(T0//3I) Relays remained essential for many 
devices. The army put together increasingly pow- 
erful versions of the "Joos polygraphic counter" 
using the SIS's versions of the Letterwriter equip- 
ment. Arlington Hall's men also lashed together 
machines such as the Kryha Decipherer and the 
Longitudinal Differenc ing machine. The navy 
helped with the desktop | 

a ■' ' ' 

devices, and it went much further by putting new 
types of relays to work in its electrical crib-tester 
for the Purple system, the Purple Dudbuster. 74 

(U) Relays and plugboards were used in larg- 
er, stand-alone machines. Most of them were 
quite powerful, although some were bizarre com- 
binations of primitive components. The need to 
create instant, yet reliable, crypt- 
analytic firefighters justified their 

CE S// ST) One crisis machine 
that had an almost vulgar look 
was the embodiment of a quite 
elegant statistical attack. The 
machine was the SIS's Dragon. It 
took almost a year to wire togeth- 
er all its components. 

(TS//SE) Dragon was the crib- 
dragger SIS sent to England in 
late 1944 to help GC&CS attack 
the German Tunny enciphering 
machine. It was made of four 
large racks of relay panels, rows of 
switches, a tape reader, a large 
control panel, and a set of indicat- 
ing lamps. A crib was set upon a 
plug-board, then the known set- 

tings and motions of Tunny wheels were entered 
through 200 switches. The message was read in 
by a standard tape reader. The crib, usually six to 
ten "letters/ was tested against the message. 
When the machine calculated that a significant 
match (psi patterns) had been made, it stopped. 
The lamps indicated the Tunny wheel positions. 

CE S //SI) The Dragon proved so useful that 
GC&CS built a second version after the American 
Dragon ate itself up. The noisy machine, which 
contained more than 2,000 cross-point relays, 
had a tendency to consume all the electrical 
power at GC&CS and to wear out its real contacts 
within a few weeks. 75 Despite its faults, it seemed 
such an achievement that it was sent back to the 
states after the war and was proudly displayed in 
the SIS museum. 76 

( JS//8B Advanced statistical powers were 
added to the Dragons. If the war had not ended, 
the Dragons might well have evolved into 
machines whose sophistication rivaled the 

TraVCl. ) Dragon 


Page 141 


(U) The Other Purples 

(U) In some ways the battle against the 
German military code and cipher systems was 
easier than the one against Japan's navy. The 
Germans' reliance on the Enigma allowed a con- 
centration of cryptanalytic effort. Japan was not 
as cooperative. Its military did not use either a 
single code or a single encryption machine. 
Although much was shared among Japan's sys- 
tems, the American and British codebreakers felt 
they had to start from the beginning with each 
new code, cipher, or communications subdivision 
system they encountered. 

(T0//DI) Despite the earlier successes against 
the Purple enciphering machine and the insights 
gained from prewar work on the Japanese navy's 
JN25 operational code, OP-20-G never had a 
secure entry into any of the major Japanese sys- 
tems. 78 

(TB//SI) Some of the Japanese naval systems 
yielded secrete more valuable than those gained 
from Purple, but that information came at a great 
cost. Even the rather old-fashioned code-plus- 
additive systems were as or more difficult to enter 
than Japan's high-level diplomatic machine. 
Japan's habit of frequently changing code and 
additive books compounded "G's" problems. 

( TS/ffii ) The Japanese navy used 184 systems 
between December 1941 and the end of the war. 
And some 1,000 different ciphers appeared on 
them. 79 Triumphs in uncovering the underlying 
logic of systems did not ensure they could be 
read. An American cryptanalytic victory could be 
reversed overnight by a change in additives or, 
more permanently, by the introduction of new 
code books. 

(TS//SI) Although the navy had its RAM 
machines, tabulators, and the special "NC" 
machines to help discover the structures of the 
code and cipher systems, some of the most 
important ones remained unreadable until the 

latter months of the war. One of the reasons why 
OP-20-G hurriedly constructed so many relay, 
plugboard, and Letterwriter combinations in late 
1944 and 1945 was the need to immediately 
exploit the systems that had taken so long to 

(U) But even in 1942, the least resistant and 
quickest technological path was the most rational 
choice for a search for an Ultra in the Pacific. Only 
when the tried-and-true technologies proved too 
slow was there an attempt to use digital electron- 

(U) In certain instances, avoiding the risk 
involved in applying electronics to the Japanese 
problems led to machines that should have been 
named after Rube Goldberg's inventions, rather 
than after snakes or jewels or flowers. The man- 
date to stay ahead of possible changes in the 
famous Japanese fleet operational code system, 
JN25, the rush to exploit Japan's long unreadable 
strip-cipher system, and the search to identify 
mysterious naval cipher machines led to some of 
World War IFs most unusual cryptanalytic 
machines. Viper, Python, Gypsy, Opal, Mamba, 
and their relatives were clever throwbacks, but 
still throwbacks. 

(U) New Guys and Old Guys, New 
Techniques and Old Insights 

(U) After OP-20-G and GC&CS had reached 
an agreement about the Enigma problem in early 
fall 1942, the old-timers at"G," such as Mrs. 
Driscoll, were transferred back to the more famil- 
iar Japanese puzzles. The German naval problem 
essentially was turned over to Engstrom and his 
team of bright but young and cryptanalytically 
inexperienced "outsiders." "G's" leaders, such as 
Joseph Wenger, thought that with all the mathe- 
matical skills in "M," with its advanced machines, 
and with help from the British cryptanalysts, 
their weak background in practical attack on sys- 
tems could be overcome. ** 

Page 142 

top s e eneTOeo iv ii H Tf/Rr.L to uoa, auo, gam, odr and mzl/ixi 


(T0//0I) The great American responsibility, 
the Japanese cryptanalytic problem, was given to 
"G's" professional cryptanalysts and the naval 
officers who had worked on the Pacific tasks 
before the war. They were in a separate subgroup, 
"GYP," 81 which maintained a rather well-calculat- 
ed distance from the young men working under 
Howard Engstrom. There was another group in 
Hawaii which continued on with its earlier work. 
The British frequently gave cryptanalytic advice 
and sent intercepts to both centers and to the 
cryptanalysts working in Australia. 

£ T$//SI) Washington was where cryptanalyt- 
ic research and machine design were conducted 
for both European and Asian problems. But being 
in one center did not lead to agreement about 
crypto-methods or hardware. Those working on 
the Japanese problems had different perspectives 
from the young men in the "M" group. The dis- 
tance between "M" and "P" included approaches 
to cryptanalysis. "P" was not as enthusiastic about 
abstract mathematical methods and RAMs as was 
"M." "P's" cryptanalysts did accept the reality that 
codebreaking had become amass production 
operation, but they wanted tried-and-true meth- 
ods and machinery, such as "differencing," the 
IBM equipment, and simple and reliable electro- 
mechanical analog. 82 

(T3//SI) Those preferences led to delays in 
the design and construction of advanced 
machines for the Pacific crypto-wars. "Ps" reluc- 
tance to turn to new technologies was only one 
reason for the delays. And it was only one of the 
reasons for the many failures in the Pacific 
machine project. Technological barriers and the 
difficult nature of the attacks against additive 
code systems led to a string of well-intentioned 
projects that were unable to produce machines of 
the stature of the Enigma Bombes. 

which thought, like Hooper, that advanced statis- 
tical methods were cures for all crypto-problems, 
and the old hands in "P," who trusted their long 
experience with Japanese cryptologic systems. 

(TC//SI) An example of the results of the dif- 
ferent approaches to codebreaking came in fall 
1943 when frustration over the inability to even 
identify a new Japanese system led to a blunt 
exchange between the two types of 
cryptanalysts. 83 The frictions had been building 
for some time, with "M's" newcomers more than 
suggesting that on a routine basis messages 
should be run through the IC, Comparator, Tessie 
machines, and the advanced tabulators to identi- 
fy "busts" or misuses of systems. By October 
1943, suggestions about such routine "clinical" 
procedures to find "non-random" behavior in 
new systems were turning into recriminations. 

(T6//6I) There was a suggestion in late 1943 
that unread systems always be given to a new 
group separate from the "YV cryptanalysts. The 
proposed "GO" section would guarantee the 
application of all the new statistical methods to 
the recalcitrant traffic, no matter its country of 
origin or its underlying subject matter. It would, 
of course, take control over new systems away 
from those who were immersed in operational 
cryptanalytic and codebreaking attacks against 
families of systems. 

(TS//SI) The suggestion for setting up "GO" 
did not have to wait long for a response. 
Commander L. W. Parke, one of the more experi- 
enced men in the organization, replied with a 
rather guarded criticism of the proposal. But it 
included more than a hint that it was "cryptana- 
lytic experience" rather than "casual observation" 
that usually led to solutions. Parke soon became 
much more direct. 81 

(U) A Matter of Machines and Control 

f TO/yOB The differences over cryptanalytic 
methodology led to some frictions between "M," 

The author of the subject paper shows such a 
lack of comprehension of what goes on in t'iY 
thaL it [sic] does not deserve serious considera- 
tion.. .As to the variation from random idea [sic] 


Page 143 

HMkU. i J Wii ii-wiiiimrnri -mum .Mr^.M «bb A H[1 ^| m 

a far more profitable attack is to be found in 
successive trials of known and probable meth- 
ods of cnciphcrmcnt. In other words cryplana- 
h lie experience cannot be supplanted by casual 
observers [sic] armed with machines and/or 
degrees of higher learning. The subject report is 
typical of the ideas that have come from (tM's 
eryptanalytieal research group during the past 
year. Although a few ideas have been useful, 
they were not worth the time spent in trying to 
help GM personnel to a belter under-standing of 
the problems involved in solving .Japanese 
naval ciphers. 

( TS//SI) A rejoinder to Parke's criticism of 
the intellectuals reached the desk of one of the 
old-timers who, apparently, decided it was best to 
mark the top of each page of the renewed propos- 
al for "GO" with the words "GM BURN BEFORE 

(TS//8I ) More than pride and turf wars were 
involved in the tensions between the operational 
and analytic types ofcryptanalysts. At times the 
differences in approaches led to serious interpre- 
tive problems. The now relatively well-known 
battle between the analysts in the Pacific and 
those in Washington over the JN25 additives was 
only one of several conflicts. 85 A quite similar one 
took place in the summer of 1942. 

(TS//SI) The JN39 Japanese Merchant 
Ship/Navy additive system was introduced in 
August 1941 and was broken by the cryptanalytic 
group in Hawaii. 86 But understanding the system 
did not mean instant success. The recovery of the 
critical additives was progressing very slowly. 
Then Washington decided that it would apply its 
machines and analytic techniques to the problem. 
It used its own method of machine "differencing" 
to generate six times as many additives as Hawaii 
was producing by hand. Washington insisted its 
additives were correct.* 7 

£RS//fl ' f ) When code and additive books were 
finally captured, it was learned than some three- 

quarters of the Washington machine-produced 
additives were incorrect. A retrospective on the 
problem found that the pure analytic and 
machine methods were too simplistic and that 
"speed" of processing was no substitute for such 
vital activities as making sure that the recoveries 
actually led to readable traffic. 88 The discovery 
about the results of Washington's machine attack 
on "39" was linked to similar problems with its 
work on JN25. 

(U) Such incidents perhaps restrained the 
operational cryptanalysts from requesting that 
the "M" sections engineers create advanced 
machines to help them break into the Japanese 
codes and ciphers. Asa result, most of the 
machines built for the "P" group followed in the 
tradition of Purple, a direct analog to aid the 
decryption process after a system had been solved 
through traditional techniques. 

(U) The Snake That Died Too Young, Viper 

(U) Among the many frustrations "G" had to 
endure was the struggle against what was thought 
to be one of the most important Japanese cipher 
systems. In late 1942 it appeared that the 
Japanese navy might be on the verge of introduc- 
ing a new cipher machine - one for the most 
important naval officers and ships. It had the 
potential to become another Purple for the 
Americans, a single machine-based system that 
would be relatively easy to read once the nature of 
the enciphering machine was established. The 
Americans called the system and machine Jade. 

(U) There were, perhaps, some overly hopeful 
fantasies atOP-20-G that Jade would soon 
replace the Japanese Navy's important opera- 
tional code, JN25. That additive code was being 
read, but with great effort and much worry. OP- 
20-G always fretted that the next change in it 
would be the one that permanently locked out the 

Page 144 



(U) JN157 (Jade) first appeared in December 
1942. "G" could tell from messages on other sys- 
tems that it was carrying very important high 
command communications. But the frustrations 
created by the unsuccessful attack on another 
Japanese cipher machine, Coral, led "G" to con- 
clude that Jade was unbeatable. Until mid-1943 
only minimal attention was paid to JN157 

{ ¥//FOUO) ~Then there were a series of very 
lucky discoveries. Some busts were identified 
which gave a few clues to the nature of the Jade 
machine and which pointed to sources for cribs. 90 
The belief that it was perhaps a solvable tele- 
phone stepping-switch machine (like Purple), 
and an increase in the amount of traffic on the 
system, led to a major attack on Jade. 91 

( 3 S //0I) * G tatistical attacks were supplement- 
ed with crib-based "menuing." "G's" engineers 
lashed together a primitive stepping switch ver- 
sion of a new bombe to aid in the search for daily 
settings of the machine. 92 With the help of such 
machines, the attack on Jade yielded results with- 
in a few weeks. 

(U) By October, 4,000 messages were being 
read by the Americans each month. The messages 
contained much about logistics and, later, many 
intelligence items. The cryptanalysts at"G," 
impressed by the high-level addresses on the 
messages, anticipated that Jade would soon carry 
the most important operational orders and would 
become more significant than the unpredictable 

(T0//SI) The optimism about the possibility 
of conquering Jade had led to the speedy creation 
of the Viper, an electromechanical analog of it, 
and the construction of several handy crypto- 
aids. 93 To speed decryption of all the messages, 
several more copies of Viper were built. The later 
models were quite advanced and expensive desk- 
top automatic decipherers. Two copies were sent 
to England. 

(U) Viper was in the tradition of Purple, 
although Lawrence Steinhardt was its top design- 
er. A special Kana electric keyboard was connect- 
ed to a large bank of electric stepping switches, 
relays, and plugboards. A Letterwriter typewriter 
was at the other end and served as the printer for 
the system. The Vipers looked much like the later 
versions of the Purple analog. The Vipers saved 
hundreds of very precious hours of analysts' time. 
But they were special-purpose machines. 

(U) Because of the apparent value of JN157, a 
major project was begun to create more than an 
analog of the system. OP-20-G put the groups at 
NCR to work designing an ultra-high-speed 
"grenade" to speed the final and most difficult 
steps in discovering the keys to the Jade setups. 
The machine was to be powerful enough to over- 
come changes the Japanese might make to the 

(U) Unfortunately, although the Americans 
continued to penetrate the JN157 Jade system, it 
did not carry the expected high-level operational 
messages. By early 1944, "G's" investment was 
not paying great returns. Then one of the great 
disappointments of the war occurred. 

(U) Jade turned out to be an experiment by 
the Japanese, one that did not please them. The 
Jade JN157 system was cancelled in mid-1944, 
just as more advanced cryptanalytic machines 
to attack the system were being delivered to 
OP-20-G. 94 

(U) It seemed that more than a year and one- 
half of intense work had been wasted. In retro- 
spect, however, it was decided that Jade had not 
provided that much important information. So its 
closing was not that critical from the operational 
side. And the work on it did make something of a 
secondary contribution to OP-20-G. It gave the 
cryptanalysts hope that another system which 
had resisted the most sophisticated attacks for 
many yearns might finally be conquered. On the 


Page 145 


basis of the work on Jade, the Coral system proj- 
ect was restarted. 

(U) A Snake in Hand, Perhaps - Python 

(U) Japan had placed naval attaches in many 
of its embassies after World War 1. Their job was 
to report on the capabilities and intentions of for- 
eign navies. 95 The Japanese hierarchy valued 
their communications so much that the attache 
system was given one of the nation's first encryp- 
tion machines, the Red. With the onset of World 
War II, the attaches assigned to the Axis nations 
assumed an expanded role in Japan's intelligence 
system. If their messages could be read by the 
Allies, they would provide insights into the plans 
and the technical prowess of all of the Axis pow- 

(TS//SI) The attaches had many ciphers to 
communicate with (eighty-three over the course 
of the war), but two seemed of special importance 
to the British and American cryptanalysts, JNAio 
and JNA20. They appeared to be the ones that 
carried the most valuable information to and 
from the attaches. The first was a very difficult 
version of Japan's many code-plus-additive sys- 
tems; the other, the "G" and GC&CS teams dis- 
covered, was a new type of cipher machine that 
would not respond to the attacks that had broken 
Red or Purple. They called their new adversary 

(TS//SI) Through increased attention to the 
problem in Washington and England, the begin- 
nings of an entry into JNA10 (the code and addi- 
tive system) came in late 1943. However, full 
reading of all the links between the attaches and 
Tokyo was delayed for many months because spe- 
cial additive books had been assigned for com- 
munications between foreign capitals and Japan. 
They proved even more resistant to statistical 
attacks than the standard additives. 96 

(T8//SI) There was elation when the fortu- 
nate discovery of some cribs led JNA10 to open 

up to the Allies. The entry came too late to be of 
value, however. JNAio's main use was for trans- 
mission of information gained through espi- 
onage. By 1944 Japan's spy networks had with- 
ered, and the system yielded little information. 

(U) However, the attack on JNAIO did give 
some encouragement to those who had attempt- 
ed to read the much more important system, 
JNA20-Coral. JNA20 had frustrated "G" for 
many years. The navy needed a successful attack 
against it because it was used to report important 
technical information concerning all the military: 
land, air, and sea. As significant, its transmissions 
came from such vital points as Berlin and 

GES// 8 9 JNA20 had a long, long history, but 
the Coral machine had first appeared in 1939 
replacing the Red machine. It was soon distrib- 
uted to all of Japan's attaches. The revised JNA20 
system immediately came to the attention of the 
British. Although unable to read the traffic, they 
continued to intercept messages throughout the 
war. OP-20-G also took notice of the 1939 change 
in JNA20, but it was unable to intercept enough 
of the transmissions from Europe to Japan to 
begin an attack. That put the Americans at a dis- 
advantage compared to the British codebreakers; 
in addition to the shortage of manpower, the lack 
of intercepts prevented the Americans from 
acquiring the necessary cryptanalytic "depths." 97 

(U) The desire to penetrate Coral remained 
high. Because OP-20-G had been reading Coral's 
predecessor, the Red machine, its officers real- 
ized how significant the Coral was. As many men 
and machines as could be safely taken off of other 
systems, such as the one for Japanese naval oper- 
ations, were assigned to the new attache problem 
in 1940. 

C T B //BI) "G" tried all the standard statistical 
approaches. The messages that were available 
were frequency analyzed to see if Coral had the 

Page 146 



same kind of statistical split between consonants 
and vowels as the old Red and Purple machines. 98 

(TC//SI) Sadly, there was no return from the 
investment before the summer ofi94i- Even 
then, the results were slim and less than encour- 
aging. The time-consuming statistical tests 
showed the analysts only that Coral's inner work- 
ings were not just slight variations on the old Red 
machine. Coral, they knew, would be a very 
demanding problem. 

(U) The task seemed daunting, so resources 
were shifted to other problems. However, when 
workloads allowed, "G's" analysts were encour- 
aged to apply the most advanced methods to 
uncover the fundamentals of Coral. Some signs of 
progress appeared. They led to the reestablish- 
ment of a Coral program. 

(T3//3I) Many different statistical analyses 
and all the available RAM machines were applied 
during the next two years. The new IC machine 
was put to use, then Tessie. Some 100,000 mes- 
sage letters were counted and matched, again and 
again. Digraph counts and strip studies were 
done to see if Coral was related to the Enigma. 
Index of Coincidence tests were made on five 
months' worth of intercepts. Round Robin tests, 
matching every message from a day with every 
other, were made. A few prewar "obtained" plain 
texts were also analyzed. Isomorphic runs were 
done, and searches for tetragraphic repeats were 
made using all the messages for July 1941. All 
types of IBM tabulator indices were produced. 
The British joined in with their own methods, 
applying Turing's advanced hypothesis-testing 

(T S //SI) The work in England and America 
went on throughout 1942. But all that was deter- 
mined was that Coral might be employing tele- 
phone stepping switches, as did other Japanese 

(T0//GI) The attack stopped again. The RAMs 
and statistical techniques had let OP-20-G down. 
It seemed fruitless to devote any more effort to 
JNA20, despite the apparent value of its mes- 
sages. The situation seemed hopeless. There was 
little possibility of either a theft of a machine 
(they were all in enemy territories) or a major 
misuse of the system. 

(U) The victory over JNi57-Jade in fall 1943 
led, however, to renewed interest in Coral. Based 
on what was being learned about Jade, there was 
a hunch that Coral was a Roman-letter version of 
the JN157 machine. As soon as possible, the Jade 
team was reassigned and ordered to beat Coral." 

g flg// SI } They joined with their counterparts 
in England on a six-month major statistical attack 
that again called on all the RAM machines. The 
goal was to reconstruct the wiring of the Coral 
stepping switches. The work was intense, but lit- 
tle came from it. By spring 1944 many were ready 
to declare that Coral was unbeatable. 
Mathematics again seemed unable to fulfill 
Hooper's promises. 

(TS//SI) Then, perhaps to the embarrass- 
ment of the mathematical types, a new member 
of the Coral group decided to retry some cribs. He 
asked a translator to see if he could place them, 
expecting not to hear from him for days, if not 
weeks. To everyone's surprise, the translator 
returned within a hour with a report of success. 
Then a longer crib was placed and the Coral 
wirings were recovered. 100 In early March 1944 
Coral was beaten and began to yield intelligence 
treasures. 101 

(U) "G" had already been preparing for the tri- 
umph over Coral. The "M" group had been 
ordered to start designing an analog well before 
the break occurred. The result, the Python, was a 
relatively crude, quickly built desktop cousin of 
the Viper. Two Letterwriter typewriters were the 


Page 147 


input and output mechanisms. Between them 
was a plugboard and a large bank of relays that 
could be set to imitate the "wheels" and motions 
of Coral, 

(U)) Python was put to use in mid-1944, auto- 
matically decrypting messages and helping to test 
possible solutions to Coral settings. Several more 
were built and served until the end of the war. 
While relatively slow, the Pythons did their job 
with the reliability that could not be expected 
from more advanced and costly electronic 
devices. It was soon helped by a special analysis 
machine that had originally been built for the 
abandoned Jade system. 

( 3S//SI) Supported by an ongoing statistical 
attack suggested by the British, and by a home- 
grown crib method that broke daily keys, Coral 
was made to yield some of the best intelligence of 
the war. The cargo-carrying submarines that ran 
high-priority material between Japan and 
Germany became easy targets. The German 
"buzz" bomb, radar, and other great secrets 
became known to the Allies through the reports 
sent to Japan by the attache in Berlin. 

(U) That attache did more for the Allies. He 
gave a precise description of all the defenses on 
the French coast. In addition, his messages told 
the Allies what the Germans thought about the 
timing and location of the Allied invasion of 

(U) Of Strips and Stripper 

(U) Japan did not trust cipher machines as 
much as Germany did; it employed several rather 
old-fashioned systems on many of its important 
communications channels. Some of those primi- 
tive systems caused as much trouble or more than 
Enigma or Jade or Coral. One of Japan's alterna- 
tive encryption devices, JN87, led to as much 
technological soul-searching atOP-20-G and 
National Cash Register as had the cipher machine 

(U) The Americans had suspected that the 
Japanese navy was going to initiate the use of a 
new shipboard encryption system in mid-1944. 
But beyond a hint that it would be a sort of hand- 
operated "strip" device, nothing was known. 
When the Japanese began using the system, K G" 
was unable to read any of its messages. Then a 
capture was made by Filipino troops. The instruc- 
tion books and parts they seized in November 
1944 were quickly sent to "G." The captures and 
analysis of JN87 intercepts led to a partial but 
important solution by early December. 

(U) JNS/s device was quite like the American 
Navy's own strip cipher. The '87 had aplastic 
board holding strips that had alphabets printed 
on both sides. There was a stock of one hundred 
two-sided strips to chose from. Thirty at a time 
were placed in the board, with their particular 
vertical and horizontal arrangement set accord- 
ing to complex specifications given in a book of 

(TB) The JN87-based fleet communications 
became so numerous that hand deciphering was 
an impossibility. Thousands of intercepts began 
to pile up; the backlog seemed to contain items of 
importance. To solve the problem, "G" instructed 
Lawrence Steinhardt to build a reliable machine 
as quickly as possible. His design reflected how 
little time he had to complete the analog. At least 
he was able to show a sense of humor: he gave the 
machine the name of the famous American strip- 
tease dancer, Gypsy Rose Lee. 102 

4S) The NCML's engineers at NCR were given 
the responsibility for construction. They worked 
on the control portion of Gypsy, a formidable 
task. Ralph L. Palmer was in charge, and he was 
determined that his team would overcome all the 
difficulties. His crew worked double shifts and 
stayed on the job through the Christmas holidays. 
That allowed them to solder the required 40,000 
connections in Gypsy's -central cabinet. 103 
Meanwhile, the navy's engineers and, perhaps, 
some WAVES were wiring the many "strips" 

Page 148 



(plugboards) as quickly as the cryptanalysts could 
recover them from analysis of JN87 depths. 

(T S // 8 I) Both teams were successful. Gypsy 
was in operation a month after it was ordered. 
Unfortunately, just as Gypsy began its work, the 
Japanese altered their system, and the boards 
had to be rewired. The Washington unit became 
very efficient at that; the Japanese began to 
change all the strips every three months or so. 104 

(T0//i> r) Steinhardt's Gypsy was aget-the- 
job-done machine. It was a large, 4,000-pound 
relay, stepping-switch, and plugboard 
combination that required a central 
control unit and five separate six-foot- 
high bays. Each of the bays contained 
five large plugboards. Each board was 
hand-wired to represent four of the 
JN87 strips. Because the strips were 
two-sided, the Gypsy plugboards were 
constructed to represent eight choices. 

(TS//SI) The control unit had 
switches to select the desired "strips" 
and which of their sides were to be 
used. The Gypsy's operator could also 
automatically shift the "strips" and set 
other crypto-variables. Banks of lights 
signalled which boards and which off- 
sets had been selected. Then the enci- 
phered message was typed on the 
machine's Kana input typewriter. If the 
message was free of garbles and Gypsy 
was correctly set, clear text was printed 
on a Letterwriter typewriter. The JN87 
messages seemed so important and 
Gypsy so useful that another model was 
ordered. It was to be used in Hawaii. It 
was planned to have a tape reader as 
input to speed processing. 

(U) Strips without Strippers 

happy career. Topaz and Mamba became exam- 
ples of the danger of relying upon special-purpose 
machines, even when they could be built quickly 
with tried-and-true parts. Both Topaz and 
Mamba arrived after their target problems had 

(U) Topaz was quite similar to Gypsy in its 
architecture and purpose. It was another huge 
plugboard and relay combination for semiauto- 
matic deciphering through removal of the influ- 
ence of "strips." But its task was a bit different 
than Gypsy's. The "strips" used by the Japanese 

£¥&) Two other machines built for 
"strip" problems did not have such a 

(IO//OI ) Manfca 


Page 149 

top ocora:T//oo Mi NT//rc[:L, to uoa, auo, caw, qbr an p nzu/xi 

on their navy merchant marine JNn communica- 
tions represented additives. JN11 was a superen- 
ciphered code system and a very important one 
which often served as the window into Japanese 
communications when other codes were unread- 

(U) Topaz was clever. It removed the additives 
and printed the resulting code. But the two mod- 
els of Topaz were rarely used. They arrived too 
late to be employed on current JN11 traffic, and, 
despite their being modified to work on the 
famous JN25 system, their short lives were spent 
working on old JN11 messages. 107 

($} One of the most unusual machines that 
OP-20-G brought into its RAM collection was 
Mamba. It combined IBM cards, relay circuits, 
and analog decision making in a five-foot-high 
metal cabinet. It looked like a friendly, many-eyed 
monster because of its two round signal lights and 
its small voltage meters. ,o8 

i&) Mamba's unique feature was its input sys- 
tem, an electromechanical card scanner. Located 
on a shelf at the center of the machine, the scan- 
ner shifted IBM cards holding cipher over a set of 
others punched for key. Mamba sensed how 
many of the 2,400 tiny metal brushes (ten digits 
in eighty columns in three cards) made contact as 
the message cards were moved, column by col- 
umn, over the key cards. Mamba's goal was to 
exploit a weakness in many of the Japanese code 
systems, one that had become important to the 
American attacks. All legitimate code groups were 
evenly divisible by three. 

rB> Mamba's electrical weighting component 
could be set to stop the machine when accumu- 
lated "scores" exceeded a threshold value. When 
enough of the brushes made contact through 
coincident holes in the cards, the scanner's power 
was cut off. Once the machine stopped, its opera- 
tors recorded the source of the "hits" by hand. 

-6S) Although relatively crude, and useful only 
when the cryptanalysts had a good idea of which 
page of additives had been applied to a message, 
Mamba seemed so promising that two were con- 
structed at NCR. They were delivered to the navy 
in the last days of 1944. 

(U//rOUO) It was expected that Mamba 
would vastly reduce the amount of hand work 
needed for the JN11 system. But the Mambas 
arrived after the specific system they were 
designed for "died." To salvage some of the 
investment, at least one was modified to perform 
what was called a "maximal-minimal" attack on 
JN25. 109 

(U) The Attack on the Many JN25S 

(U) The new automatic enciphering 
machines, such as the Enigma, Jade, and Coral, 
have been crypto-historians' favorites. The 
machines and the attack on them have an inher- 
ent attraction. The mechanical complexities are 
fascinating, and the logic of the attacks are intel- 
lectually challenging. In addition, the machines 
that were built to fight them, the Bombes and the 
Colossi, are symbols of the coming of the new 
information age. The Bombes and Colossi are also 
physical reminders of how difficult codebreaking 
had become by the 1940s. 

(U) But more important in terms of the war in 
the Pacific than the automatic enciphering 
machines were old-fashioned systems, such as 
the Japanese Fleet General-Purpose Code, and 
JN25. Codes like "25" proved as or more difficult 
to penetrate than machine ciphers. JN25, an 
additive system, took so much effort that a special 
and very large "G" cryptanalytic group was creat- 
ed early in the war. 

(U) In Washington alone, some 800 people 
were working on "25." "° By the later months of 
1943> Howard Engstrom's "M" group became 

Page 150 



involved. It began to devote much of its time to 
finding practical designs for very, very rapid 
machines with enormous amounts of memory to 
handle the special demands of the JN25 problem. 

(U) Because of the special challenges posed by 
code systems, "G" was unable to create an equiv- 
alent to the Bombe for the Japanese codes. And 
they were unable to devise electronic devices. "M" 
had to concentrate on creating "memory" 
machines from what was technologically at hand. 
Although compromises, the machines they came 
up with promised to be more productive than the 
standard tabulators, the navy's NC machines, or 
the relay-tabulator Slide Run machines that were 
used by the army." 1 

•£¥&) JN25 was a difficult problem for several 
reasons. The primary one was the general nature 
of well-fashioned code plus additive systems. 
They left analysts awash in unknowns, forcing 
them to grab at cryptanalytic straws to make any 
progress. Although codes can be analyzed 
through frequency tests quite similar to those 
used on the distributions of letters in cipher sys- 
tems, doing so is quite demanding and usually 
yields less rewarding answers. Counts of code 
groups in large collections of messages can yield 
good pointers to frequently used words in a lan- 
guage, such as "to" or "the," and to word combi- 
nations typically found in military messages such 
as "Fleet orders." 

fPS)- Identifying one group can help find the 
meaning of another. But codes with thousands of 
groups, several of which stand for the same plain 
language word, usually do not open up as a result 
of frequency tests. 

(U) The frustrations involved in trying to 
solve a code system without the aid of captures or 
operator errors grow exponentially when the 
underlying codes are themselves enciphered. 
Adding or subtracting numbers tends to mask the 
frequencies of the code groups. If the "additives" 
are from a long list or, worse, are randomly gen- 

erated, there is little chance of removing their 

(U) Additive systems have been in use around 
the world for generations, and attacks on them 
began when they first appeared. Some very com- 
plex methods had been developed to help identi- 
fy additives. Attacks such as "differencing" were 
relatively freestanding. To begin the arduous 
process of rounds of subtraction and cycle analy- 
sis, differencing only asked for messages that 
were in "depth." But much depth was required. 
Differencing needed bulk cipher-text that was 
known to have been enciphered with the same 
key. But the method had at least the potential for 
automation. 112 

(U) In contrast to differencing, most other 
approaches were based on common sense, were 
difficult to automate, and were useful only after 
some significant breakthroughs into the additive 
and code systems had been achieved. A frequent- 
ly used one was to assume that some additives 
were known, then subtract them from the cipher, 
then match the resulting plain code against a long 
list of known and highly frequent code groups. If 
a match, or a significant number of matches, was 
found, additional effort was invested in produc- 
ing a plain text and increasing the list of known 
code groups. 

(U) Other methods demanded the use of very 
experienced codebreakers. They required knowl- 
edge of military systems and a craftsman's insight 
once additives had been removed. For example, 
an analyst might begin his attack by assuming 
that previously known stereotyped phrases or 
usages were in the messages. After seeing if his 
intuition led to an interpretation that made sense, 
a tentative meaning would be assigned to a code 
group. Then other messages would be searched to 
double check the hunch. If consistency was 
found, many messages would be processed in 
order to construct a file of appropriately weighted 
"high frequency" code groups. That file would be 
checked and updated in an endless cycle as it was 

top ocorcrn/oo MiN TOrccL to uca, auc, ca n , o b r and n zujxi 

Page 151 


used to see if correct additives had been discov- 
ered, and to see if meanings could be attached to 
more of the recovered numeric codes. 

(U) The Americans had used all those meth- 
ods in the 1920s and 1930s, and they applied 
them to Japan's World War II systems. But JN25 
had its own particular difficulties, ones that chal- 
lenged the traditional attacks. First, it had a very 
lengthy code book. Its first one, of 1939, con- 
tained over 30,000 groups; later versions had 
more than 50,000 of the five-digit codes. That 
meant that even when a clear code group was 
intercepted, its meaning was very difficult to 
determine. More than one code number might 
stand for the same word or phrase. In addition, 
the Americans thought they might be facing a sys- 
tem in which one code group could have different 
meanings depending upon context. 

(U) None of the old cryptanalytic standbys 
appeared in JN25's early years. Unfortunately for 
"G," JN25 code books were not "obtained," there 
were few operator mistakes, and OP-20-G was 
unable to discover many cases in which a JN25 
message was sent on a system that was being 
read. The only recourse OP-20-G had was to try 
to intensify the application of the tedious and 
frustrating traditional attacks it had used before 
the war. 

(U) The Japanese did not make things easy 
for "G," although they did employ a few tech- 
niques that deviated from the rule of random- 
ness. One of the errors the Japanese made, but an 
important one, was to facilitate the detection of 
garbled messages by making legitimate code 
groups divisible by three. 113 The use of that "pat- 
tern" vastly reduced the number of true code 
groups the Americans had to identify, and it also 
eased the search for additives. 

(U) The Americans had achieved some bril- 
liant but temporary successes against JN25 using 
old-fashioned hand methods supplemented by 
the IBM tabulators. Enough of JN25 was under- 

stood in mid- 1942 to give "G" its first great tri- 
umph. 114 With only a handful of people and a few 
tabulating machines, "G's" branch in Hawaii was 
able to contribute vital information on the com- 
ing Midway and Coral Sea battles. The contribu- 
tion was marked by controversy, however. 

(U) The analysts in the Pacific had disagreed 
with the conclusions by "G's" experts in 
Washington about the correct interpretation of 
the codes and additives. The analytic group in 
Hawaii had faith in "experience" while 
Washington relied more upon formal analytic 
methods. Unfortunately for the chief analyst in 
Hawaii, Joseph J. Rochefort, being correct about 
the contents of JN25 did not prevent him from 
being disciplined for deviating from prescribed 
reporting practices. 113 

(TS//SI) But there was much in the JN25 sys- 
tems that made them very, very difficult. The 
codes were hidden by the use of nearly random 
additives that were contained in lengthy books. 
The first of the additive books had 300 pages, 
each holding 100 five-digit numbers. A complex 
and very opaque indicator system was employed 
to communicate which page and which starting 
point were to be used for a message. All JN25 
indicators were scrambled in one way or another, 
and their encryption algorithms were frequently 

(U) More than indicator systems were altered 
to thwart cryptanalysis. During the first two years 
of its life, the JN25 code was replaced once and 
the additive books were changed six times. The 
replacement of the additive books continued 
throughout the war. Worse, the code itself was 
changed at very critical moments. That meant the 
need to reconstruct the code meanings all over 
again. When the Japanese changed additives, 
codes, and indicator systems all at the same time, 
it was devastating for the Americans. That was 
the reaction in 1944 when there was a total 
change in "25." 

Page 152 



■ (T0//0D Despite all the experience it had 
gained, in mid-1944 "G" worried that it might 
never reenter JN25. Only errors on the less valu- 
able JN11 systems allowed it into operational 
messages while its crew frantically tried to deter- 
mine the meanings of the thousands of new "25" 
codegroups. 116 

(U) The difficulties caused by the JN25 sys- 
tem changes were multiplied as the Japanese 
divided the "25" system into more and more sep- 
arate networks or "channels." Each had its own 
procedures. Entry into one was not a guarantee 
that others would be read. 

(V) The Comparators That Weren't: the 
Copperhead Proposals and the Victory of 

£S} It had become clear to OP-20-G and the 
SIS, which was tackling similar problems, that 
automation was needed to handle the Japanese 
codes. The first responses were predictable. Both 
the army and navy turned to mechanics, tabula- 
tors, and relay devices. The SIS engineers and 
their allies at IBM extended the reach of the relay 
tabulator combinations through the Slide-Run 
machines. "G" counted on similar devices until 
"M" began to involve itself with JN25. It recom- 
mended the construction of a host of permuta- 
tions of Bush's ideas. Unfortunately, only one was 
accepted, and it evolved into a punch tape, not a 
microfilm machine. 

(S)-The navy had asked the practical engi- 
neers at the Navy Yard under Don Seiler for help 
fighting code systems before the war began. By 
1940 he built a fifty-wheel cam and gear device 
that was driven by hand cranks. It tested cipher 
and additives for divisibility by three. 117 The next 
response by both the army and navy was to mod- 
ify the tabulators. The goal was to speed up the 
false subtraction processes used in both differ- 
encing and additive stripping. The navy's NC 
machines began arriving before the war broke 
out, and the army quickly drafted plans for its 

Slide Run machines. Those combinations of com- 
plex relay boxes and tabulator equipment were 
difficult to build, however, and did not come into 
operation until the last months of 1943. " 8 

£S} OP-20-GM's team wanted to create 
machines faster and much more powerful than 
the NCs or the Slide Run devices. Lawrence 
Steinhardt was given the responsibility. He 
helped draft a series of proposals in mid-1942. 
Each fit with his and John Howard's previous 
work at MIT. While the designs were being 
mulled over, Steinhardt rescued the idea that film 
and photoelectric technology could be used for 
frequency tests against JN25. Apparently such a 
hope had almost been killed in early 1942. 119 

(S3 One of the first challenges Howard 
Engstrom had presented to Eastman-Kodak was 
to produce a machine for additive code systems. 
With the ex-IBM engineer John Skinner acting as 
the liaison, Eastman quickly arrived at an ambi- 
tious design for an "automatic decoder." The 
device was to strip additives, locate the plain code 
in a large dictionary, and set down the meaning of 
the code. Eastman was not sure whether elec- 
tronics or older technologies would be used to do 
the stripping; but it was committed to developing 
optical discs to store the code meanings and to a 
fast-flash system to print the results. To "G's" dis- 
appointment, just as Engstrom thought of asking 
Eastman to prepare a detailed design, the compa- 
ny announced it was too busy with the IC and 
Tessie projects. 12 ° 

{&} Given the difficulties at Eastman, 
Engstrom turned to Lawrence Steinhardt. 121 He 
handed him the responsibility for the additive 
systems. He began exploring machine alterna- 
tives in mid-1942. But he was not allowed the 
opportunity to turn his general ideas into 
specifics until mid-1943. The Atlantic problem 
took all of "M's" resources until then. 

(S) Steinhardt's initial suggestions relied 
upon Bush's favored technologies. They were 

™ "T""rTfi lnn l"" Tlinn T" "1 ii * i , /WE, "AN, ^RR flMn N7I f l 

Page 153 


sometimes referred to as the Copperhead propos- 
als. The first of them was for a photo-optical 
machine that was to be a high-speed replacement 
of Seller's device. It was to have base-three tube 
ring counters. Its job was to subtract suspected 
additives code groups found in messages already 
aligned in depth. Its goal was to point to the most 
likely additives through the "divisible-by-three" 
criterion. It took a year after the initial sketch to 
produce a detailed design. Then when it was 
examined, it was abandoned, declared to be too 
electronically adventuresome. 

TS^The next plan, also envisioning photo- 
optical technology, was more architecturally 
ambitious. It was intended to automate the 
sophisticated method of'Jeeping." It was a 
means of identifying likely additives and then 
code groups. "Jeeping" was an extension of the 
differencing method; an extension that called for 
a large amount of high-speed memory and near- 
endless rounds of running messages against each 
other. It was based on the probability that if 
cipher texts that had been enciphered with the 
same "keys" were subtracted from each other, an 
identifiable "difference" would lead to plain code 
and correct additives. The difference could be 
checked against a huge catalog containing all the 
possible differences between high frequency code 
groups in a system. "Jeeping" was a valuable 
method, but it soaked up hours of effort to 
process just a few "differences." 

#f) Although "Jeeping" was a method that 
should have been automated, the photo-optical 
machine, with the high-speed photographic 
reproduction of hits that it demanded, was too 
much for "M." The detailed mid-1943 pian for the 
proposed Mark II was shelved. Even a simplified 
version, without automatic recording, was aban- 

(S^Bush's fundamentals received another set- 
back when Copperhead II was rejected as too 

complicated. It was a film version of a Slide Run 
machine. Its job was to subtract known additives 
in a system from masses of cipher text, then run 
the results against a huge file of already known 
high-frequency plain codes. If enough matches 
resulted, the machine was to issue a signal and 
record the hit. With that information, an analyst 
could align more text and add to the files of 
known additives and codes. 123 

fS^Even a reduced version of Copperhead II 
was rejected. The Mark Vwas to have a smaller 
"memory," although its matching decisions were 
to be based on a complex weighting system 
derived from studies of language and code fre- 
quencies. The debates over the merits of those 
"Hall" and "Shinn" weights may have been one 
reason why the Mark V was not completed. 

(S}-The only one of Stein hardt's designs that 
was accepted and turned into hardware was 
Copperhead I. And its technology was a major 
compromise. It became a punched tape, not a 
film machine. That it was built at all is an indica- 
tion of how much JN25 worried the cryptana- 
lysts. 124 

4$)- Copperhead I was a device to aid them 
when they had no entry into a system. It was an 
embodiment of a "Brute Force" method that was 
used in moments of desperation. Its purpose was 
to fill the void when the cryptanalysts did not 
have enough recovered additive or code values to 
even begin using tools such as differencing or 
"Jeeping." A large volume of cipher text was run 
in the hopes of finding "double repeats" which 
would indicate which messages and their offsets 
were in depth. A double repeat was when the 
same two encrypted code values appeared in two 
texts at the same distance apart from each other. 
It was calculated that locating such matches vast- 
ly increased the probability that a depth had been 
found. 125 

Page 154 



(U) Beyond the Copperheads - the JN25 
Crisis and "M's" Response 

tT0//0t ) Howard Engstrom would have 
ordered Lawrence Steinhardt to turn toother 
problems after the Copperhead defeats, but there 
were signs that JN25 and other Japanese additive 
systems were undergoing a series of changes. 
There was fear they might become unreadable. 
High-speed machines to meet the new Japanese 
challenges seemed essential. As a result, 
Steinhardt was ordered to explore all types of 
alternatives to Bush's favored technologies. 

(TS//SI) As Steinhardt searched for new pos- 
sibilities, including digital electronics, other engi- 
neers at SIS and "M" moved towards electronic 
solutions to German challenges. Like Steinhardt, 
they were pushed by cryptanalytic needs that 
could not possibly be met with the older tech- 
nologies. By 1944 both the army and navy were 
moving far beyond the original Bombes and 
Bush's Comparator. 126 


1. (U) Bradley F. Smith's The Codebreakers War, 
(Novato: Presidio Press, 1993) details the struggles of 
the army to gain entry into Ultra. 

2. C F S//SI) NSA CCH Series XII Z, "Washington E 
Traffic, Notes on Correspondence" circa February 
1942. (TS//SI ) NSA CCH Series XII Z, "Green Analog," 
May, 1953. A most important source showing the con- 
centrated work SIS did on the complex German diplo- 
matic systems is found in ( TS//SI ) NSA CCH Series 
XII Z, "History, Machine Branch," np. nd. (T0//0I) 
NSA CCH Series IVB-1-11, "History of the Signal 
Security Agency, Volume 11, The Machine Branch," 
October 29, 1947. 

3. (T0//0I) NSA AHA ACC 16844. "History of the 
Special Projects Branch, SIS ETOUSA." 

4.(U) NSA release, Theodore M.Hannah, "Frank 
B. Rowlett: A Personal Profile," 522. NARA RG457, 
SRH-004, "The Friedman Lectures on Cryptology." 

5. (U) Thomas Parrish, The Ultra Americans: The 
United States' Role in Breaking the Nazi Code (New 
York: Stein & Day, 1987), 45- 

6.(U) NARA RG457, SRH-361, "History of the 
Signal Security Agency," Volume II, 82, and SRH-362, 
"History ofthe SSA Vol. Ill, The Japanese Army- 
Problems: Cryptanalysis, 1942-1945." Edward J. Drea, 
MacArthur's Ultra (University of Kansas Press, 1992), 
10. NARA RG457, SRH-145, "Collection of 
Memoranda on Operations of SIS Intercept Activities 
and Dissemination 1942-1945," 01, and SRH-361, 
"History ofthe SSA" Vol. II, 250, 272. Ronald Lewin, 
The American Magic, 38 (New York: Farrar-Strauss, 
1982). (S//SI) NSA CCH Oral History Interview 
OH 04-82 with Samuel S. Snyder, 24 February 1982. 

7.(U) NARA RG457, SRH-004, "The Friedman 
Lectures on Cryptology," 171. 

8. (U) Cipher A. Deavours and Louis Kruh, 
Machine Cryptography and Modern Cryptanalysis 
(Dedham, Massachusetts Artech House, 1985), 238. 
NARA RG457, SRH-305, "The Undeclared War: The 
History ofRI," 15 November 1943, byLaurance F. 
Safford, Captain, U.S. Navy, and SRH-159, 
"Preliminary Historical Report ofthe Solution ofthe B 

9. (U) Edward J. Drea, MacArthur's Ultra 
(Lawrence: University of Kansas Press, 1992), xii, 

10. (U) Again, the documents found in the Garland 
Covert Warfare series are most rewarding. See 
"History of3-US," 010-026, and "Origins, Functions 
and Problems ofthe Special Branch, MI." Useful back- 
ground on army intelligence is in Bruce W. Bidwell, 
History ofthe Military Intelligence Division, 
Department ofthe Army General Staff: 1775-1941, 
(University Publications of America, nd). 

11. r9) NSA RAM File, June 23, 1943, OP-20-G to 
OP-20, "Army has agreed to tell England too much." 
NARA RG457, SRH-349, "Achievements ofthe SSA In 
World War II," 31, and SRH-361, "History ofthe Sig- 
nal Security Agency, Volume Two, The General 
Cryptanalytic Problems," 11-22, 250, 276-283. 

"TS^NSA CCH Series XII Z, "History of GET (TUNNY) 

12. (U) Edward J. Drea, MacArthur's Ultra 
(Lawrence: University of Kansas Press, 1992), xii. 


Page 155 


Geoffrey Ballard, On Ultra Active Seruice (Rich- 
mond, Australia: Spectrum Publications, 1991), 194- 

13. fS^ Perhaps it was SIS's first failure with Freak 
that led Joseph Desch to accept the design for Mike, 
the huge mechanical counter NCR built later in the 
war. fSfNSA CCH Series XII Z, Inventories of RAM 
Equipment, 1945. 

14-^SJ- Perhaps it was SIS's first failure with Freak 
that led Joseph Desch to accept the design for Mike, 
the huge mechanical counter NCR built later in the 
war, f[oo58]. 

15. 0SO//OI) NSA CCH Series XII Z,"MA.C. 
Outlines #11, Freak." &$). NSA CCH IX.B.1.9, SSA, 
"History of the Signal Security Agency, Volume Nine, 
History of the Development Branch," 10 February 


16. (U) The use of condensers for such storage 
was "in the air" at the time, including at MIT, and sim- 
ilar systems were used in the ENIAC. 

17. (TS//SI) The seven condensers could hold 128, 
but Freak counted only up to 99. pTG//3fr NSA CCH 
Series XII Z, "M.AC. Outlines #11, Freak." 

18. #>•) Many of the storage uses of condensers 
were based upon setting constant values by hand. 
Thus, Freak was quite an adventure. 

19. fg.) NSA CCH Series XII Z, "Freak I," May 1953. 

20. e») NSA CCH IX.B.1.9, SSA "History of the 
Signal Security Agency, Volume Nine, History of the 
Development Branch," 10 February 1953, 83. 

21. fl S //8q . NARA RG457, SRH-349, "Achieve- 
ments of the SSA In World War II," 18. University of 
Pennsylvania Van Pelt library Archives, Papers of 
John Mauchly, 2B-io:a 209, 14, October 11, 1945 and 
April 14, 1945 "Visit to SIS and Cryptologic Problems." 
NARA RG457, SRH-361, "History- of the Signal 
Security Agency, Volume Two, The General 
Cryptanalytic Problems," 237. (TB//Sf ) NSA CCH 
Series XII Z, "History, Machine Branch," np. nd. 
(■ EG//93 NSA CCH Series IVB-1-11, "History of the 
Signal Security Agency, Volume 11, The Machine 
Branch," October 29, 1947. 

22. (TB//SP NSA CCH Series XII Z, "History, 
Machine Branch," np. nd., 30-37. 

23. (3S) NSA CCH Series IV, V 10.6, Chief Signal 
Officer, "A Chronology of the Cooperation Between the 

SSA and the London Office ofGCCS," 2 June 1946. 
(TB//CI) NSA CCH Series XII Z, "Washington E 
Traffic, Notes on Correspondence" circa February 
1942. ( TC//OI) NSA CCH Series TV B-i-u, "History of 
the Signal Security Agency, Volume 11, The Machine 
Branch, October 29, 1947. 

24. (TD//OI ) NSA CCD Series IVB-1-11, "History 
of the Signal Security Agency, Volume 11, The Machine 
Branch," October 29, 1947, 84. 

25. (TS//& I) NSA CCD Series IV B-i-n, "History of 
the Signal Security Agency, Volume 11, The Machine 
Branch," October 29, 1947, 22-23, 88. (TO//OI) NSA 
CCH Series XII Z, "History, Machine Branch," np. 
nd., 24. 

26. tm NSA CCH Series IV B, "History of ASA 
Equipment (Development Branch) History," 
December 1942, 30 June 1944- 

27. ■{TS//SI) ft 42211] ( TS//DI) NSA CCH Series 
XII Z, "History, Machine Branch," np. nd., 51, 62. 

28. OS#Si) NSA Series XII Z, MAC Outlines, 
"The Slide Run Machine." 

29. ff 0/ / 8f) NSA Series XII Z,MAC Outlines, 
"The Slide Run Machine." (TS//SH NSA CCH Series 
IVB-l-ii, "History of the Signal Security Agency, 
Volume 11, The Machine Branch," October 29, 1947. 
The first "F"-built machines had two cabinets. 

30. ( TO//Or)- NSA CCH Series XII Z, "Office of 
Computers, List of Computers," nd. Other relay-tab 
combinations such as the JMA and the "deciphering 
machine" were built by SIS to perform similar tasks for 
additive systems. 

31. (TS//SI) NSA CCH Series XII Z, MAC Outline 
#4, "The Slide Run Machine." 

32. CTO//3I) NSA CCH Series II Z, MAC Outline 
#4, "Slide Run Machine." 

33- m NSA CCH Series XII Z.0P-20-G, "SSA 
Proposal for 70mm Film I. C. Machine," 8 June 1945. 

34. (TS//SI ) NSA CCH Series TVB-1-2, "History 
of the Signal Security Agency, Volume Two; The 
General Cryptanalytic Problem," 272. 

35. Cm NARA RG457, SRMA011, "Senior Staff 
Meeting Notes," August 18, 1942, Friedman memoran- 
dum "Establish Section F." David J.Crawford, The 
Autoscritcher and the Superscritcher, forthcoming. 
The Annals of the History of Computing illustrates the 
advanced technical achievements of"F." In fact, "F" 

Page 156 



may have forged a bit ahead of OP-20-G in respect to 
the use of digital electronics. One reason may have 
been that "F" was under less pressure to solve imme- 
diate cryptologic crises. Again, NSA SRH-391, "U. S. 
Cryptologic History," contains dates somewhat differ- 
ent than those found in RAM file documents and 
other relevant SRH volumes. fP9> ASA CCH Series IV 
B, "History of ASA Equipment (Development Branch) 
History, December 1942 -30 June 1944-" (S//SI) 
NSA CCH Oral History Interview OH 04-82 with 
Samuel S. Snyder, 24 February 1982, 96. 

36. (U) NARA RG457, SRH-361, "History of the 
Signal Security Agency, Volume Two, The General 
Cryptanalytic Problems," 287. 

37. ( ¥ 8 //S i) NSA CCH Series IV B-1-2, "History of 
the Signal Security Agency, Volume Two: The General 
Cryptanalytic Problem," 272. 

38. ffS//0r) NSA CCH Series IV B-1-2, "History of 
the Signal Security Agency, Volume Two: The General 
Cryptanalytic Problem," 258. 

39. (U) NSA SRH-391, "U. S. Cryptologic History," 
120, provides a hint that the SIS's relay machine may 
have first been explored by the British and shown to 
the Americans under direct order from Churchill. 
Other sources, such as a letter from George Stibitz to 
the author, suggest the relay machine was an 
American idea. 

40. (TS//SI) NSA RAM File, J. N. Wenger to 
OP-20-G, September 3,1942, "Part II of Report of 
J. N. Wenger, Capt. USN," 1. Letter to the author from 
George R. Stibitz, June 7,1987. NSA RG457, 
SRH-361, "History of the Signal Security Agency," 
250, 257, 272-3. At least one source claims that the 
American army's cryptologists were informed of and 
worked on the FISH traffic as early as August 1942. 
That source also claims that some machines were built 
in America for the automated solution of that binary 
system. However, there is no claim that the army built 
anything like the Colossus for the problem. See NARA 
RG457, SRH-349, "Achievements of the SSA In 
World War II," 18. 

41. £S} Rosen did not lose faith in electronic solu- 
tions. In June 1943 he suggested a machine solution 
for the commercial Enigmas that would include a 
frame from the relay Madame X bombe and as many 
as fifty-two "counters" made of vacuum tubes. 

(^-NSA CCH Series XII Z, Robert o. Ferner, "Rapid 
Analytic Machinery Needed for Research," June 3, 


42. (T3//3I) NSA CCH Series IV B-1-2, "History of 
the Signal Security Agency, Volume Two: The General 
Cryptanalytic Problem," 258. {¥9* NSA CCH Series IV 
B, "History of ASA Equipment (Development Branch) 
History, December 1942-30 June 1944. f[422i]. The 
contract with Western Electric was for only about 
$2,000. The F branch annual report for 1944 listed a 
project for a high-speed rotor for the period May - 
November 1944. $8) NSA CCH Series XII Z, Annual 
Reports, Development Branch, 1943-4. 

43. (TS//OI) NARA RG227, Box 73, February 29, 
1944, Stibitz toNDRC, "Secrecy re NCR product." 
Williams went on to build many huge relay computers 
for the military ordnance groups during World War II, 
and he designed and patented an electronic computer. 
Michael R. Williams, A History of Computing 
Technology (Englewood Cliffs, New Jersey: Prentice- 
Hall, 1985), 225-240. Hagley Museum and Library, 
Accession 1825, Honeywell v Speiry-Rand , Trial 
Records, February, 1942, S.B. Williams toNDRC, 
"Fire control proposal," and Reports on Electronic 
Computer Designs byS. B.Williams, November 1941, 
January 1942, March 13, 1942. NSA RAM File, 
September 3, 1942, "Wenger to OP-20-G, bombe 
project;" September 9, 1942, "Machine Research 
Section (F)"; "Part II of Report of J. N. Wenger, Capt. 
USN," land October 10, 1942, "Enigma Machine 
Contract." Letter to the author from George R. Stibitz, 
June 7, 1987. The army's single machine cost over 
$1,000,000. NSA RG457, SRH361, "History ofthe 
Signal Security Agency," 257, 272-3. NARA RG457, 
SRH-349, "Achievements ofthe SSA In World War II," 
29. NARA RG457, SRH-61, "History ofthe Signal 
Security Agency, Volume Two, The General 
Cryptanalytic Problems," 251. 

44- (TS//SI) NSA CCH Series XII Z,0P-20-G 
"Memoranda on Bombe and the relationship ofthe 
U. S. and U. K.," circa 1943. In addition to not fulfill- 
ing all the expectations of automatic setups, 003 did 
not incorporate some features suggested after its first 
design was set. In mid-1943 one ofthe SIS cryptana- 
lysts suggested that a Madame X frame be combined 
with electronic counters to allow a purely statistical 

top or.onr.T/tt)O MiN T</nr.L to uoa, auc, gam, oon a n d hzl//xi 

Page 157 


attack. A later machine, the SIS Dudbuster did have 
something like that configuration, but it became a sep- 
arate machine, not an integral part of 003. (9)- NSA 
CCH Series XII Z, Robert 0. Ferner, "Rapid Analytic 
Machinery Needed for Research," June 3, 1943. 
Technical details of 003 may be found in an early 
report, (S) NSA CCH Series XII Z, "X-68003, Bell 
Laboratories Report, April 2, i943- " 't 3//3I) NSA CCH 
Oral History Interview OH 04-82 with Samuel S. 
Snyder, 24 February 1982.. 

45. (TS//SI) After some experience with the 003, a 
menu with a sure crib of twelve letters was considered 
useful although it would produce more false hits than 
a strong menu of fifteen or so letters 

46. (S//CI) NSA CCH Series XII Z, Oral History 
Interview OH 04-82 with Samuel S. Snyder, 24 
February 1982. 

47- (TD//0I) NSA CCH Series XII Z(S-2568), 
"Tentative Brief Descriptions of Cryptanalytic 
Equipment for Enigma Problems," circa 1945. 

48. ( - ¥S//S1) NSA CCH Series IV B-1-2, "History of 
the Signal Security Agency, Volume Two: The General 
Cryptanalytic Problem," 265. (*8) (S2568) NSA CCH 
Series XII Z, "Tentative Brief Description of General 
Analytic Equipment for Enigma Problems," 26 March, 
1945- W NSA CCH Series XII Z, "Cryptanalytic 
Report #2: The Yellow Machine," 3, 54. 

49. (TS//6I) NSA RAM File, February 21,1944, 
W.A Wright toOP-20-G, "Comparison of Army and 
Navy Enigma Equipment," and January 18, 1943, to 
OP-20-G/da, "Report of Meeting on Army Bombe." 

50. iSy NSA CCH Series XII Z, "Cryptanalytic 
Report #2: The Yellow Machine," 50-52. 

51. (T0//0I) NSA CCH Series IV B-1-2, "History of 
the Signal Security' Agency, Volume Two: The General 
Cryptanalytic Problem," 265. For insight into how 
much SIS had learned about Enigma methods by the 
end of 1943, see (6) NSA CCH Series XII Z, 
"Cryptanalytic Report #2: The Yellow Machine," 
especially pages 31 and 35. These suggest that the 
British had developed their own "machine guns." 
( T3//3I) NSA CCH Local Archive, "Army-Navy 
Descriptive Dictionary of Cryptologic Terms," Army 
Security Agency, February i947--fS)- NSA AHA ACC 
16890N, "Bombe Operations, Control, and Testing, 
Duds and Railwav E." 

52. ( TS//BI) NSA CCH Series IV B-1-2, "History of 
the Signal Security Agency, Volume Two: The General 
Cryptanalytic Problem," 265rf9)-NSA CCH Series XII 
Z, "Cryptanalytic Report #2: The Yellow Machine," 35. 

53- W NSA AHA ACC 16890N "Bombe 
Operations, Control, and Testing, Duds and Railway 

54- (T3//3I) NSA AHA 16331, "6812th Signal 
Security Detachment (PROV) Apo 413 Army," 
15 June 1945, 35- 4® NSA CCH Series XII Z, 
"Cryptanalytic Report #2: The Yellow Machine," 40. 

55- ( T0//0I) NSA CCH Series IV B-1-2, "History of 
the Signal Security Agency, Volume Two: The General 
Cryptanalytic Problem," 265. 

56. (TS//SI) NSA CCH Series XII Z, LeRoy H. 
Wheatley, "Cryptanalytic Machines in NSA" 30 May 
1953, and various years. 

57- -GS} The SIS cryptanalysts, Ferner and Small, 
worked on several statistical and technological 
approaches to a "Dudbuster." -$S) NSA CCH Series XII 
Z, "Cryptanalytic Report #2: The Yellow- Machine." 
f&} NSA CCH Series XII Z, Robert 0. Ferner, 
"Rapid Analytic Machinery Needed for Research," 
June 3, 1943. 

58. (TS//SI) NSA AHA ACC 13657, "G.C. &C.S. 
Naval SIGINT. Vol III, German Cryptographic 
Systems and Their Solution," 204. 

59. (T3//3I) NSA CCH Series XII Z, copies of 
various MAC Outlines, circa 1953, MAC Outline # 12, 
"The Arlington Dudbuster." 

60. rS) NSA CCH Series XII Z, Robert o. Ferner, 
"Rapid Analytic Machinery Needed for Research," 
June 3, 1943- 

6i. fTS//SI ) NSA CCH Series XII Z, "History of the 
Signal Security Agency, Volume Two, The General 
Cryptanalytic Problems," 266. r¥S> (S2568) NSA CCH 
Series XII Z, "Tentative Brief Description of General 
Analytic Equipment for Enigma Problems," 26 March, 


62. (¥ S//SQ NSA CCH Series IV B-1-2, "History 
of the Signal Security Agency, Volume Two: The 
General Cryptanalytic Problem," 266. 

63. ffO//S» NSA CCH Series XII Z, LeRoy H. 
Wheatley, "Cryptanalytic Machines in NSA" 30 May 
1953, and various years. 

Page 15S 



64. fl ' S//5T ) NSA AHA ACC 13657, "G.C. & C. S. 
Naval SIGINT, Vol 111, "German Cryptographic 
Systems and Their Solution." 

65. CSS.) (S2568) NSA CCH Series XII Z, 
"Tentative Brief Description of General Analytic 
Equipment for Enigma Problems," 26 March, 1945. 
t&rNSA CCH Series XII Z, "Cryptanalytic Report #2: 
The Yellow Machine," 52. 

66. (U) NSA RAM File, Part II of Report to J. N. 
Wenger, Capt. USN, "Resume of the Dayton, Ohio 
Activity During World War II," and "History of NCML 
and OP-20-G-4E, June, 1944," "n530 bombes in oper- 

67. £PS) NSA AHA ACC 35701 "History of the 
Bombe Project," 16 February 1946. 

68. (T3//3I) NSA CCH Series XII Z, (S12OO8) 
Navy Dept., Office of Chief of Naval Operations, 
DNC(0P-20-G), RIP 425, "The American Attack on 
the German Naval Ciphers," October 1944 [sic]. On 
Britain's request to SIS to build a Bombe and to build 
analog of the "G" machine, •£$) ASA CCH Series IV B, 
"History of ASA Equipment (Development Branch) 
History, December 1942," 30 June 1944, 42-44 

69. PSJjSf) NSA CCH Series XII Z, "History of 
the Signal Security Agency, Volume Two, The General 
Cryptanalytic Problems." (T0//3f) NSA AHA ACC 
16844. "History of the Special Projects Branch, SIS 
ETOUSA"tSTNSA CCH Series XII Z, "Cryptanalytic 
Report #2: The Yellow Machine," 44. 

70. {¥S)_ NSA AHA 35529, Friedman to 
Corderman, 29 March, 1944, "Comparison of our 
"003" type of "Bombe" with the rotary type." 

71. (S) NSA CCH Series XII Z, Annual Reports, 
Development Branch, 1943-4. 

72. (U) F. H. Hinsley, British Intelligence in the 
Second World War, Volume I (New York: Cambridge 
University Press, 1979), 58. NARA RG457, SRH-361, 
"History of the Signal Security Agency, Volume Two, 
The General Cryptanalytic Problem," 15, 243, 269, 277. 

73. (U) David J. Crawford, The Autoseritcher and 
the Superscritcher, forthcoming, The Annals of the 
History of Computing , NARA RG457, SRH-361, 
"History of the Signal Security Agency, Volume Two, 
The General Cryptanalytic Problems," 269-270. 

74.-f rO//aij MAC Outline 30, MAC Outlines 103, 
SATYR. 4& NSA AHA ACC 26373, Chief, "F" 

Branch, "RAM Equipment," 29 March 1945. The first 
model of Satyr was built at Dayton in late 1944. Four 
additional models were made by the navy and the 
SIS group built its own version. All of them were quite 
direct analogs of the popular Hagelin machine. They 
even incorporated wheels from an actual Hagelin. 
Relays and plugboards eased the task of setting the 

75. fei NSA CCH, G. 0. Hayward "Operation 
Tunny: Deciphering German Teleprinter Traffic in 
WWII atBletchley Park," 14 July 1989, 
Z/1396GW/9000/5, 9- 

76. (T3//3I) NSA CCH Series XII Z, CNO, CIT 
Paper TS47, "Report on British Attack on FISH," 
Washington, May 1945. f/ES) NSA CCH Series XII Z, 
"Fish'Dragon Notes," February 1945. (TS//SB NSA 
CCH Series XII Z, "Fish Notes," 17 January 1945. 
( T3//3I) NSA CCH Series XII Z, "MAC. Outlines #21, 
Tunny Dragon." 

77- (TS//SI) NSA CCH Series XII Z,"MA.C. 
Outlines #21, Tunny Dragon." 

78. ( TS//SI) NSA CCH Series IV.W.I.5.13, 
"The History of OP-20-GYP-1, 1939-1945." The very- 
serious and frightening blackouts caused by changes 
to Japanese systems, especially the trauma of fall 1944, 
are well described in (TS//SI) NSA CCH Series 
IV.W.1.25.12, "General History of OP-20-3-GYP," 
new Chapter 111, espec. 24. 

79- ( TS//S Q NSA CCH Series IV.W.1.5.12, 
"History of OP-20-3-GYP," new Chapter III, 2. 

80. (■¥87YSTTNSAL-566o CCH Series IV.W.1.5.12, 
"General History of OP-20-3-GYP," Appendix 1,5. 

81. ( TO//0I) Most of the following discussion 
refers only to the group that was based in Washington 
and which had a designation of GYP-i. ( ¥0//9 t) NSA 
CCH Series IV.W.1.6.8, S163287, "The History of 
GYP-i." The cryptanalytic or"Y" section ofOP-20-G 
subdivided several times and its bureaucratic history is 
quite complex. 

82. (TS//SI ) NSAL-5660 CCH Series IV.W.1.5.12, 
General History ofOP-20-3-GYP, Appendix 1,7. 

83. ff S//M ) NSA CCH Series XII Z, "File Kept by 
Dr. Campaigne on Ram Panel Meetings," Ely memo- 
randum 3 October 1942 and attached memoranda 
dating to 1 November 1943. 


Page 159 

TOP oconerffooMi N Tf/ncL to uoa AUO, CA N , QDR AMD NZLW 1 

84. (W/») NSA CCH Series XII Z, "File Kept by 
Dr. Carapaigne on Ram Panel Meetings," 1 November 
*943> "Further Remarks on the Proposed Clinical 
Attack." The quotations are from ibid., L. W. Parke, 
"Clinical Attack on Unknown Cipher System..." 
23 October, 1943. 

85. (U) Rear Admiral Edwin T. Layton, U. S. N. 
(Ret.) etal., And I Was There: Pearl Harbor and 
Midway-Breaking the Secrets (New York; William 
Morrow and Company, Inc., 1985), 409. 

86. (U) There are some unconfirmed rumors that 
success was partially due to the theft of a codebook 
from a Japanese ship docked in San Francisco. 

87. (TS//SI ) NSA CCH Series IV.W.1.5.12., 
"General History of OP-20-3-GYP," new chpt. V, 6. 

88. fTS/61) Such problems may have led to the 
search for machines that would perform Slide Run 
type dictionary checks. 

89. (T0//S1 ) NSA CCH Series IV.W.1.5.12., 
"General History of OP-20-3-GYP," new chpt. V, 6. 
"the significant reasons for this" [the JN39 error] 
"were three (experience with similar processes in 
JN 25 show that generalizations are feasible}:...." 

90. (TS//SI) NSA CCH Series XII Z, "General 
History of OP-20-3-GYP," 

91. £B S #gI) "G" obtained a badly damaged copy 
of the machine in December 1944. NSA AHA ACC 
17480 "Final Report, Project P123, Original J.N.157 
Machine," February 1945, OP-20-G-4-D-3E. 

92.-^ NSA CCH Series XII Z, H. H. Campaigne 
"Use of Hypo on the JN-157," 21 February 1944, 
gives an insight into how the general-purpose RAMs 
were employed to attack Jade. The Hypo was vised to 
set the starting positions of Jade's three moving step- 
ping switches. Jade had three moving stepping switch- 
es, two immobile ones, and astecker. ffiSZ/ft i) 
NSA CCH Local Archive, "Army-Navy Descriptive 
Dictionary of Cryptologic Terms," Army Security 
Agency, Februarys 1947. A machine. Jasmine, whose 
details seem to have been lost, was built to test cribs 
against Jade. Avery primitive electrical and stepping 
switch device, Mortor, was quickly constructed to 
allow hand testing of cribs to see if crib-plain pairs 
completed a circuit. 

93. Q SSff&f ) The first Viper was proposed as 
soon as it was thought that Jade might be conquered. 

While it was being built, a handy but clumsy "bombe" 
for Jade was built, the Mortor. It was a cluster of step- 
ping switches and wires to handtest menus to solve 
Jade settings. It would be replaced by the Rattler, a 
machine described below. (8) NSA CCH Series XII Z, 
"Viper, Plans for Construction of, Steinhardt, L. R.," 
7 Sept. 1943- 

94 . (TS//SI) NSA CCH Series XII Z, "General 
History of OP-20-3-GYF." 

95. (TS//SI ) NSA CCH Series IVW.L5.12, 
"General History of Op-20-3-GYP." 

96. CP0//SQ NSA CCH Series IV. W.1.5.12, 
"General History of OP-20-3-GYP." 

97. (TS//0I) NSA CCH Series XII Z, "History 
of (NAT) JNA20 CORAL," Vol. Ill, and NSA CCH 
IV. W.1.5.12, "General History of OP-20-3-GYP." 

98. mffm) NSA CCH Series XII Z, "General 
History of OP-20-3-GYP." 

99. CTS//ST) NSA CCH Series XII Z, "History of 
JNA20 Coral (NAT) Volume MI." 74- 

100. (TS//SI ) NSA CCH Series XII Z, "File Kept 
by Dr. Campaigne on Ram Panel Meetings," memo- 
randum on "Personnel inGM-2," November 1943 
indicates that some significant NAT busts were also 
discovered by new or inexperienced personnel. Busts 
were also important to the entry- into JN157. 

101. (TB//6I) NSA CCH Series IXW 1.5.12, 
"General History of OP-20-3-GYP." 

102. OES.) S338 NSA CCH Series XII Z, 
L.R. Steinhardt, OP-20-G-4a-5, "A proposed form 
of Gypsy..." 13 Dec. 1944. NSA CCH XII Z, "Project 
M-312, Gypsy," 28 February 1945, NCA, Washington, 

103. W NSA CCH Series XII Z, "Viper, Plans 
for Construction of, Steinhardt, L R.," 7 Sept. 1943. 

104. (TS//ST) NSA AHA ACC 17480 "Final 
Report, Project P123, Original J.N.157 Machine," 
28 February 1945, OP-20-G-4-D-3, 5. 

105. (¥5) NSA CCH Series XII Z, "Proposal for a 
Tape Reader on Gypsy," 2 April 1945. 

106. (*S#SI) NSA CCH Series XII Z, CBO CIT 
Paper TS-31, "TOPAZ," Washington, December, 1945. 
(T0//0I) On the history ofJNn, NSA CCH 
Series IV W.1.5.12, "General History of OP-20-3-GYP," 
new Chapter ITT, 20. 

Page 160 



107. TS) NSA CCH Series XII Z, Inventories of 
RAM Equipment, 1945. The first TOPAZ seems to 
have been completed in March 1945. T?9) NSA CCH 
Series XII Z, CBO CIT Paper TS-31, "TOPAZ," 
Washington, December, 1945, 2. A smaller but similar 
machine, ASP, had an even shorter life in 1944, when 
the Japanese changed their callsign systems. 

108. -m NSA CCH Series XII Z, "Cryptanalytic 
Phases ofMamba," GM-2 6 April 1944.-^ NSA 
AHA ACC 26373, "Inventory of RAM Equipment," 
January 1945- 

109. ( TO//S4 ) NSA CCH Series XII Z, "Mamba." 
( TO/701) NSA CCH Series XII Z, "Communications 
Supplementary Activities, RAMP Report II," 
21 December 1948. (*}- NSA CCH Series XII Z, 
RAM list and Conference at Dayton, 11 April 1945. 
$}■ NSA/CCH Series XII Z, "Cryptanalytic Phases of 
Mamba," GM-2 6 April 1944. 

110. (TS//SI) NSA CCH Series IV.W.1.5.12, 
"General History of OP-20-3-GYP," Appendix 1, 7. 

111. f FS// SI) NSA AHA ACC 543 "MAC. 
Outlines #4, Slide Run Machine." 

112. (U) NSA, Lambros D. Callimahos and William 
F. Friedman, Military Cryptanalysis, Part II, NSA, 
1959, 230-238. 

113. (TS//SI) NSA CCH Series IV.W.I.5.13, 
"The History of OP-20-GYP-1, 1939-1945," 1, 
points out that the sum of the digits had to equal 
three. Divisibility meant without carry. 

114. (T0//0I) OP-20-G did make some other early 
contributions. A theft of books had allowed it into the 
Japanese merchant ship systems from 1929 through 
the summer of 1941. With the experience gained from 
that, reentry into such systems began again during 
1942. Throughout the war, the information from such 
systems, including those handled by SIS, proved of 
great significance to America's submarine fleet hi the 
Pacific. See (TS//SI) NSA CCH IV.W.1.5.12, "General 
History of OP-20-3-GYP," newchpt. V, 4-5. 

115. (U) Rear Admiral Edwin T. Layton, U. S. N. 
(Ret.) etal., And I Was There: Pearl Harbor and 
Midway; Breaking the Secrets (New York: William 
Morrow and Company, Inc.), 1985, 409. 

116. (IS#M) NSA CCH Series IV.W.1.5.12, 
"General History of OP-20-3-GYP." 

117. (Sf NSA CCH Series XII Z,L. RSteinhardt, 
"Additive Machines: Historical Summary of," 
27 November 1944. 

118. (£8) f[4233] f[4222]. 

119. tSj NSA CCH Series XII Z, L. R. Steinhardt, 
"Full-Selector," 31 October 1942. 

120. m NSA CCH Series XII Z, OP-20-GM 
J. A. Skinner, "Proposal for a Decoding Device," 
16 February 1942. 

121. *9) NSA CCH Series XII Z, OP-20-G, "SSA 
Proposal for 70mm Film I. C. Machine," 8 June 1945 
C T S/ /8 fl NSA CCH Series XII Z, OP-20-G "Analysis 
of Analytical Machine Attack onJN-37," 24 March 
1945. (TS//SI) NSA CCH Series XII Z, R. A. Rowley, 
"Preparation of Weighting Film, Secondary Stage 
Problem," Op-20-G, 2 August 1945- OP-20-G and the 
SIS would return to Eastman later in the war with pro- 
posals for sophisticated film-based machines. Two 
were turned into hardware by Eastman, the Amber 
and the 5202, before or close to the end of the war. 
They are discussed in the next chapter. 

122. (S)-NSA CCH Series XII Z, L. R. Steinhardt, 
"Additive Machines: Historical Summary of," 
27 November 1944. {S) NSA AHA 1505, John N. 
Seaman, "Memorandum for Major Edgerton, Liaison 
with Navy # 3, Use of Ramon Jap Naval Problems of 
BII Type," 9 June 1944. Note that the Gray-NCR 
Comparator was also used on the JN25 problem. 
However, it performed the required tests very slowly. 
48y NSA CCH Series XI E, Hagelin, Box 2, Folder, 

123. (8) Steinhardt, L. H., "Copperhead II (Project 
M-230) Final Report," 9 November 1944. This a 
fascinating description of the proposed design. It was 
to have a long film tape with the text and additives 
run against each other. As they did so, the subtraction 
process would yield a "mask." The known groups 
would be on another long film. As its contents were 
projected against the mask, photocells would register 
how much light passed through. They would trigger 
electronic counters which, once reaching a threshold 
value, would indicate the position of hits. 

124. 4& NSA CCH Series XII Z, L. R. Steinhardt, 
"Additive Machines: Historical Summary of," 
27 November 1944, indicates that the NCR-built 
special desktop electromechanical machine was 


Page 161 

top ocora:T//ooM i NT//ra:L to uoa, auo, cam, qdr a n d n zu/x i 

a result of the "additive" problem design effort. 
Some sixty of those Mark IV or "fruit" or "Big Adam" 
machines were constructed. Tf*}- NSA CCH Series XII 
Z, OP-20-G, "Additive Theory, Folder IV, Miscellaneous, 
Part A," 1942-43, gives details on a hand-held system 
using IBM cards (6,000 to a set) as stencils to visually 
identify divisible by three code-additive combinations. It 
also contains "G's" ideas for other types of additive 

125. 4ft NSA CCH Series XII Z, OP-20-GH-F, 
13 November 1943, Steinhardt, L. R, "JN-25 Double 
Pentagraph High Speed Machine for Locating.'"(S}.NSA 
CCH Series XII Z, OP-20-G, "Additive Theory, Folder 
IV, Miscellaneous, Part A," 1942-43, points out that 
Japanese errors frequently played a role in allowing "G" 
to recover enough additives to begin analysis. But it 
argues that with high-speed machinery, analysts would 
no longer have to depend upon them. 

126. (SS) OP-20-G-4-A5, 23 November 1944, L. R. 
Steinhardt, "Possible Engineering Solutions for 
Full Selector Problems." (TS//SI) NSA CCH Series 
IV.W.1.5.12. "General History of OP-20-3-GYP," 
New Chpt. Ill, 18, 23. The JNII and JN25 crises also 
led to proposals for "quick-fix" mechanical devices, 
See f¥Sr) NSA CCH Series XII Z, L. R. Steinhardt, 
"JN-11 (George Molecular Attack) Machine Aid For," 
27 July 1945. 



Chapter 6 
(U) Beyond the Bombes and Beyond World War II 

(U) Some of the cryptanalytic emergencies the 
British and Americans confronted pushed them 
to create machines that were close to being com- 
puters; at least the machines contained hints of 
the great potentials of electronic calculation. But 
despite the wish of many of the young army and 
navy electrical engineers to "show their stuff' and 
create the most advanced machines possible, the 
pace of innovation was determined by cryptana- 
lytic needs rather than by electronic visions. "G" 
and "F" were arms of operational agencies, not 
research organizations. The two groups built 
some of the most complex electronic computing 
machines in the world during the war, but their 
duty was to solve problems rather than invent 
perfect automata. That led them away from seri- 
ous consideration of either a universal program- 
mable machine or a binary-based computer. 

(U) Perhaps that was a wise decision. Those in 
America who had committed to an attempt to 
create universal machines saw their projects yield 
quickly outdated technological patchworks. For 
example, Vannevar Bush's Rockefeller Analyser 
was a conglomeration of electronic, electrical, and 
mechanical components that was put to rest soon 
after the war ended. The Harvard-IBM project 
under Howard Aiken depended upon the crafts- 
man's art of combining IBM card- reading equip- 
ment, relays, pulleys, and shafts. Even the 
wartime project at the University of 
Pennsylvania, which began with a commitment to 
the use of electronics, ended with a batch of spe- 
cial-purpose calculation boxes linked by huge 
cables rather than by a software program. None 
of the grand attempts created the ultra- high- 
speed and full universal machines that had been 
hoped for. 

(U) Although OP-20-G and the SIS did not 
aim for the great prize of a single computer for 

every function, they achieved a great deal. By the 
time Japan surrendered, the Americans were 
building electronic machines using twice as many 
tubes as the British Colossus. The advances in 
electronics at the cryptanalytic centers were 
amazing. But in several ways the Americans' 
achievements were limited. The cryptanalytic 
problems they solved with digital electronics were 
not memory dependent, and some of the new 
electronic machines they built were based on very 
clever ways to make analog technology imitate 
digital methods. And the new machines were not 
true data processors. Although the navy had its 
Copperheads and Comparators, large files 
remained in the domain of the tabulators and 

(U) After the Bombe 

(U) In late 1943, just as the first OP-20-G and 
SIS Bombes were being completed, another stage 
in the development of cryptanalytic machines 
began. Both American engineering groups 
returned to a consideration of digital electronics. 
At the same time, they began to pay attention to 
the Japanese problems. 

(U) The war in the Pacific was an American 
show, and the cryptanalytic work was not clut- 
tered with the kind of difficulties that complicat- 
ed the European relationships. OP-20-G and SIS 
had much more freedom, and the British were 
more cooperative. Despite the greater independ- 
ence, the Pacific never received as much attention 
from OP-20-G's and SIS's machine builders as 
did the Atlantic. There was no crash program to 
develop expensive devices to conquer the 
Japanese code and cipher systems. However, the 
engineers in Washington and Dayton put a great 
deal of work into solving problems for the crypt- 
analysts assigned to the Asian traffic. 


Page 163 


(U) Most of the work on special machines for 
the war against Japan was done in OP-20-G. And 
much of that was directed by one ofVannevar 
Bush's ex-students, Lawrence Steinhardt. He had 
been left in Washington during 1942 and 1943 to 
design what became the Copperhead tape scan- 
ning systems and to start building very advanced 
analogs of several different Japanese encryption 
machines. He was also charged with the responsi- 
bility for the machines for all major Japanese 
code systems. 

its of computing technology. Mass memory was 
the key. Lawrence Steinhardt realized that. After 
consulting with the OP-20-G cryptanalysts about 
what new methods they wished to implement, he 
returned to his superiors with an estimate of the 
probable cost for a code machine. It was high. But 
he was told that his proposed expenditure of 
$500,000 would be acceptable. That was one- 
half of what Madame X had cost and the price of 
eleven Bombes. But a solution to problems such 
as JN25 was worth many millions of dollars. 

(U) Every Which Way: The Code Challenge 

(U) The Japanese additive code problems 
challenged OP-20-G's capabilities throughout the 
war. But there were moments of urgency that led 
to bursts of activity with the army's and navy's 
engineering groups. In mid-1944, when there 
were signs that Japan might begin yet another 
series of alterations that might close its most 
important systems to the Allies, "G" intensified its 
search for methods and machines. Lawrence 
Steinhardt was again detailed to seek out techno- 
logical solutions. What he recommended indi- 
cates how deeply code solutions had become 
dependent on massive data processing. 

(TS//6I) T he cryptanalysts had no easy ways 
to solve code and additive systems. Even the most 
advanced methods of the time demanded tens of 
millions of tests and massive amounts of memo- 
ry. Probable additives had to be stripped, the 
results run against a large dictionary, and a judg- 
ment made as to whether a true code group had 
been recovered. Then meaning had to be attached 
to the clear code. There were no great mathemat- 
ical shortcuts for codebreakers. Even the most 
efficient methods called for exceptional amounts 
of labor, or powerful machines, ones that did not 
exist in 1944. 

- CES// S I) If "G" wanted a machine that could 
go beyond the army's tabulator-relay Slide Run, it 
would have to ask its engineers to stretch the lim- 

(T8//SI) Steinhardt began a survey of techni- 
cal possibilities for a machine that would allow 
"G" to employ its various new and more powerful 
versions of the additive stripping/high frequency 
tests or, if desired, the Jeeping method. This time 
he did not even bother to determine if film or 
even punch tape systems were "memory" possi- 
bilities. The sour experience with the earlier 
Copperheads and the delays in the Eastman 
Kodak film-based code device effort led him to 
explore other alternatives for the critical high- 
speed mass memory called for by the cryp- 

(TS//SI) 1 Steinhardt evaluated all the tech- 
nologies used by aspiring computer builders, 
including some that would become integral parts 
of the first modern computers. He did not discov- 
er any ready-made solutions to the high-speed 
memory problem, however. Nor could he find an 
easy solution to the challenge of constructing the 
switching system needed to select memory ele- 
ments. His frustrations grew when "G's" cryptan- 
alysts asked him to focus on a particular problem 
and to turn one of their most demanding methods 
into hardware. They asked him to design and con- 
struct a machine to attack JN25 and to do it with- 
in a few weeks. 

£TS//0I t The machine the cryptanalysts 
dreamed about for the JN25 code problems was 
an ambitious one. What was later called the 
"Selector" was to read at least 100 enciphered 
five-digit code groups at a time, rapidly subtract 

Page 164 



either additives or another set of codes, check for 
divisibility, and then perform the critical step: 
compare the resulting clear groups with a list of 
scientifically weighted code groups 1 (100,000 of 
them) and calculate whether or not a statistical 
threshold had been reached. 2 If the combined 
"weight" scores for the matched groups summed 
to or exceeded a specified level, then the device 
would signal that true additives might have been 

(T8//3I) A critical part of the required 
machine was a method of quickly changing the 
scores associated with the "dictionary" of code 
groups. The cryptanalysts wanted to modify the 
scores as they learned more about the system or 
when they desired to switch the machine from a 
weighted frequency to a Jeeping mode. 

(j¥S) With those requirements in mind, 
Steinhardt called upon his past experience at 
MIT, talked with his contacts on World War II 
computer projects, such as the one at Harvard, 
and reviewed what he had learned on earlier OP- 
20-G assignments. 

(^ftr) One of the first options he explored was 
for what seemed a wild scheme for a fast memo- 
ry. Although it had first been proposed as an 
alternative to the commutators used on the 
Bombes, the option was soon recognized as a sig- 
nificant "memory" possibility. It was a primitive 
version of what later came to be called the elec- 
trostatic storage tube, a television-like device that 
used a charged spot to hold a "bit" of informa- 
tion. 3 

rS} To follow up on the idea, in 1944 "M" had 
begun exploring the possibilities of a modified 
oscilloscope. Its beam would be electronically 
deflected to any one of several hundred spots on 
its face; then small metal patches pasted on the 
face of its screen could sense which "bits" were 
active. 4 

09} By November Steinhardt decided that he 
had learned enough about the "scope" and other 
technological possibilities and that he had to 
begin construction of a machine. He drew up a 
list of recommendations. He reported on six pos- 
sibilities for a machine for JN25 and its relatives, 
ranking them in terms of the probability they 
could be finished in time to meet the Japanese 
code emergencies. 

(U) The Navy's Madame X- the Strangest 

-(S} Although Steinhardt's report mentioned 
some very advanced alternatives, such as the 
oscilloscope memory, it argued for the use of con- 
servative technologies and architectures. They 
could "get the job done" and quickly so. As far as 
the cryptanalytic requirements would allow, 
Steinhardt wanted to use sure-fire parts and ana- 
log circuitry, but in a unique combination. 

-£S-) Steinhardt proposed a "telephone 
exchange" version of a new type of "Selector." The 
first of Steinhardt's recommendations was for the 
use of a technology the SIS had used in some of its 
machines, including Madame X, the new crossbar 
relays. With them, Steinhardt's proposed Selector 
had the potential to become one of the most pow- 
erful machines "G" or any other computer organ- 
ization ever built. 

(S> Steinhardt knew about the telephone com- 
pany's advanced relays before he went to OP-20- 
G. The late 1930s Differential Analyser project at 
MIT had used some of the "crossbar" systems the 
Bell engineers had developed for their switching 
centers. All the young MIT engineers had learned 
of impressive logical powers of the "bars." 

{&) The crossbars were miniature switching 
stations. A crossbar may be thought of as a square 
array often horizontal and ten vertical input posi- 
tions. The appropriate output is selected at the 
intersection ofthe input positions. Ifcrossbars 
were hooked together, they became powerful 


Page 165 

top ocoranv/ooM i HTy/ng to usa, aus, cam, qbr amo n z u/«i 

selectors of electrical pathways. When two cross- 
bars were connected in tandem, they could trig- 
ger the selection of one out of 10,000 switching 
paths and do it very rapidly. 

tS} Steinhardt applied his knowledge of cross- 
bars to the code-to-dictionary phase of additive 
testing. He realized that with the addition often 
small relays to a tandem setup of two crossbars, a 
five-digit code could be translated, almost 
instantly, to the electrical "address" of any one of 
100,000 locations. If the locations contained 
code groups' "weights," he reasoned, a rapid test 
for high-frequency groups might be performed. 

t9) His creativity led him much further, to the 
outline of a unique memory search methodology. 
He thought of a way to do what was, for the time, 
massively parallel "look-ups." He proposed that 
100 of the crossbar-relay combinations be linked 
together. That would allow 100 code groups to 
simultaneously link to their frequency "weights." 
It was a brilliant concept. His new Selector would 
be a parallel processor. 

fS^The crossbar provided the basis for a very 
reliable and fast digital memory Selector. But 
Steinhardt also had to find a practical way to 
match calculation speed to the rapid memory 
search. After examining electronic digital meth- 
ods of summing weights and performing thresh- 
old tests for the detection of statistically "good" 
code groups, he concluded that the most efficient 
approach was to return to the use of analog meth- 
ods and equipment. 

(U) A Wall of Knobs 

CS) The cryptanalytical method for the addi- 
tive code systems dictated a digital switching sys- 
tem to find locations of values, but it did not 
require a digital memory. Taking hold of that 
opportunity to simplify his machine, Steinhardt 
turned to an extension of previous ideas for build- 
ing high-speed memories. Electrical components 
had been suggested as means of holding constant 

values in digital form for input for calculations in 
various early precomputers. The army's Freak 
had tried to go beyond that, employing a two- 
state version based on condensers to act as a 
dynamic digital memory. That had been an ambi- 
tious and none-too-happy exercise, however. 

(&F When Steinhardt estimated the number of 
components that would be required by a digital 
memory for the weights for 100,000 code groups, 
he correctly decided to retreat to an analog mem- 
ory. If he had chosen an approach like that in 
Freak to store values as on-off representations of 
numbers, the components for the code Selector's 
memory would total to the millions. To avoid 
that, he proposed an alternative that reduced 
engineering demands. But even his clever alter- 
native called for a heroic and complex machine. 

-68} The Selector "memory" was to be a set of 
100,000 variable resistors, each with an external 
knob which was to be used to set the electrical 
"weight" for a code group. Using resistors 
reduced the number of components; only one 
resistor would be needed for each memory loca- 
tion. But even with one component per memory 
location, the memory would be an engineering 
challenge. The banks of resistors and knobs 
would have to stretch across a large room, reach- 
ing up to its ceiling. 

-$9 Steinhardts proposed resistor memory 
would be fast. But, given the amount of available 
time and manpower, Steinhardt did not plan to 
make it satisfy one of the cryptanalysts' important 
specifications. It was not to be made fully auto- 
matic. It would require a great deal of man-and- 
woman-power to set the "weights." Whenever a 
problem changed or when the cryptanalysts 
revised their list of weights, the memory would 
have to be "programmed" by resetting the 
100,000 dials. 

($} When Steinhardt first described the pro- 
posed machine option and the need to set the 
memory's values by hand, his superiors hesitated. 

Page 166 



Steinhardt admitted that it would take a crew of 
twenty WAVES a full duty watch to reset all the 
resistors. But, he argued, given the comparative 
speed and ease of construction of a resistor mem- 
ory, the eight- to ten-hour wait before a new prob- 
lem could be attacked was reasonable. Given the 
operating speed of the new Selector, a ten-hour 
setup time still left his machine with a major 
advantage over any other method of additive 
attack. 5 

(&y Steinhardt's crossbar-resistor design 
included another way of avoiding the size and 
complexity of digital electronics. The "arithmetic" 
of the machine's "frequency check" was to be ana- 
log, like the IC plate machine. One hundred val- 
ues would be sent in parallel to a circuit that test- 
ed electrical values for "enough," not how many. 

ffS//0O - The simplicity of the analog arith- 
metic circuits helped make the proposed crossbar 
machine quite fast and made its construction 
seem feasible. If the machine was set to test for 
only the weights and not strip the additives, 
Steinhardt explained, 18,000 of the 100-group 
tests could be performed in an hour. That was 
quite an advance over the army's Slide Run 
machine and the navy's NC4. And, Steinhardt 
argued, the machine could be in operation within 
less than a year because it was based on known 
technologies. 6 But he also wanted "G" to consider 
other options. 

(U) Walls of Tubes 

tS)-An inherently more attractive alternative, 
especially to a young electrical engineer, was to 
rely upon electronics. Electronic tubes, whether 
gas-filled or vacuum, were orders faster than any 
other digital technology of the time. Although 
Steinhardt believed that standard tubes could not 
be used for the Selector's memory, he thought 
they might be a possibility for the switching 
(selection) process. Therefore, his second design 
option for aJN code machine had electronic 

switching, but retained the huge resistor "electri- 
cal" memory. 

f9} As part of the JN25 project, Howard 
Engstrom had asked other "M" engineers to help 
Steinhardt by making another thorough investi- 
gation of the possibilities of electronic circuits. 
With an eye on the potential for finally creating 
an electronic Bombe, as well as building 
machines for the "weighting" attack, new tube 
technologies and circuit designs were examined. 

4£B What they reported was not good. The first 
depressing news was about the possibility of 
building an electronic wheel. The report on an 
electronic matrix which could act as a substitute 
for the Bombe commutators contained a bleak 
conclusion. With the two most reliable digital cir- 
cuit designs and standard hardware, a twenty-six 
by twenty-six matrix demanded over 1,000 tubes. 
The engineers also reported little hope for multi- 
function tubes. The many projects on radically 
new designs had not led to vast improvements. 
The available special tubes and circuits, such as 
the strobotron and Duenna circuits, still called for 
over 500 tubes per matrix. Asa result, they 
reported that an electronic selection matrix 
seemed an improbability. 

{8} The number of tubes and the likely main- 
tenance problems seemed so great that "M's" tube 
experts again turned away from digital electron- 
ics. They thought they had little chance to build 
an electronic Bombe before the war was conclud- 
ed, and they had similar thoughts about the 
chances for an electronic JN25 machine. 

"(S) They recommended another analog solu- 
tion. They pointed to an esoteric "frequency con- 
version" circuit as an alternative to the on-off dig- 
ital designs. 7 

■(SH-awrence Steinhardt did his own review of 
digital possibilities before giving the "frequency 
conversion" idea serious consideration. He put 
the electronic matrix report together with his past 


Page 167 

t op BECMrrocoMiNTf/na. to uqa, auo, qah, gbr amp mum 

experience and weighed the advantages of elec- 
tronic switching for the JN code problem. He did 
not like the results of his review, but he had to 
accept them. 

■fST His first disappointment was over the 
speed of electronics. He found that if he used a 
single matrix of tubes instead of the set of 100 
crossbars for switching and selection of weights, 
the electronic machine would be only twice as fast 
as the electric design. It would have to cycle so 
many times to find a correct pathway in the 
memory that its advantage in raw speed would be 
vastly reduced. Of course, if the single tube matrix 
was replaced with, as in the crossbar design, 100 
matrices, the electronics would make the 
machine perform not 3,600 tests per hour, but 
over 3,000,000. 

£S} That made an electronic selector very 
attractive. But such an advanced machine would 
need more than 100,000 tubes. Steinhardt real- 
ized that was too much to ask in the mid-i940S. 
Tube failures were too frequent. Based on the 
average life of standard tubes of the time, 
Steinhardt calculated that under the best condi- 
tions ten tubes would cease functioning every 
hour; by the time they were located and replaced, 
at least three more would go bad. That made the 
full electronic switch design for the proposed 
Selector unacceptable. 

4$) Steinhardt's concerns about tubes were 
based upon more than theoretical calculations. 
He had direct experience. He had worked with 
digital electronics at MIT and on the Duenna 
project at "G," 8 The Duenna project had led the 
navy's engineers to many insights on how to 
extend tube life. But even with the knowledge that 
most failures were caused by turning tube 
machines on and off, Steinhardt believed that 
unless very special types of tubes with extra long 
life were developed, 3,000 tubes were the limit 
for an operational machine. And he quite correct- 
ly saw little chance that either long life or suffi- 

ciently complex multifunction tubes could be 
developed in time to fight the Pacific code war. 9 

(U) Into the Beyond and the Past, Rooms of 
Wires and Disks 

4$) Asa result of the disappointments with 
electronics, Steinhardt took another look at older 
technologies. His survey made him more than a 
bit pessimistic about building any type of 
Selector. He had encountered some discouraging 
facts about the use of the most reliable of tech- 
nologies, standard relays. When he had calculat- 
ed how many relays would be needed to select 
and test the required 100 code groups simultane- 
ously, he was overwhelmed. Still envisioning the 
machine's memory as the collection of 100,000 
resistors and their knobs, he concluded that even 
more relays than electronic tubes would be need- 
ed for the selecting system. 

6S> A "prohibitive" number would be required 
and maintenance of such a machine, he reported, 
would be as much of a chore as keeping Madame 
X running. 

(U) Desperate Options and a Conservative 

-(8} Although Steinhardt would eventually rec- 
ommend the use of crossbars and resistors, that 
alternative was not really attractive to him. The 
thought of 100,000 resistors for the Selector's 
memory was especially troubling. So he asked 
other engineers at OP-20-G and NCR to explore 
additional possibilities. Some of the recommend- 
ed alternatives approached the bizarre. 

6S) There were last-gasp attempts to reintro- 
duce microfilm memory and suggestions for opti- 
cally read glass disks."" 

{&) There was also a brief revival of the idea of 
turning automobile parts into computers. While 
the commutators on the Bombes were distant rel- 
atives of distributors, the idea for the JN25 

Page 168 



machine suggested a much closer relationship 
between computers and automobile electrical 
systems. It was possible, some engineers said, to 
create a high-speed switching system (500 opera- 
tions per second) using ignition distributor tech- 

Jg) That suggestion does not seem to have 
been taken too seriously by Steinhardt. But 
another one that seemed to be as far-fetched did 
capture his attention. A young "G" engineer, 
Lieutenant Noble, responded with an idea that 
became the seed of one the most advanced and 
unusual research projects "G" undertook during 
the war. 

(S} Noble's idea centered upon the new and 
relatively untried technology of digital magnetic 
recording. OP-20-G had magnetic wire recorders 
that were used to copy the most important analog 
intercepts. Noble believed he could coax them 
into becoming the basis for a mass digital memo- 

(SJ He thought his proposed magnetic wire 
scanning devices could overcome the problems 
encountered with other moving media such as 
microfilm. To provide information at rates 
matching electronics, they all required such high 
transport speeds that they could not be precisely 
sensed. Despite all sorts of experiments, film, 
disk, and tape transport systems remained rela- 
tively slow and problematic. 

f$r) Noble, however, thought that he had 
found a solution, at least for wire recording. He 
thought he could line up one hundred of his rela- 
tively small wire recording devices in such a way 
that sensing difficulties would be avoided. In his 
plan, two of the differenced code digits would 
cause the switching system to select the correct 
recorder; then a sensor would select the correct 
weight as the recorder cycled through its 1,000 
values. Because each magnetic recorder held a 
few densely packed entries, processing would be 
very speedy. Resetting weights would be painless 

because the magnetic wires, Noble stated, could 
be interchanged. 12 

{8} Fortunately, Steinhardt was not forced to 
immediately choose among the many technologi- 
cal alternatives for the Selector. The JN25 prob- 
lem had eased somewhat. In addition, the mathe- 
maticians at "M" found it impossible to agree on 
which of their complex weighting schemes should 
be employed. As a result, operational cryptanalyt- 
ic attention shifted to other high-level Japanese 
naval systems. That allowed Steinhardt's team to 
avoid making any hasty technological decisions. 

- ffTS//fiI) However, they and the cryptanalysts 
decided to begin to build a experimental version 
of a new Selector. It was to be a limited four-digit 
version, almost a bread-board model. The four- 
digit version vastly reduced the potential power 
and speed of the device and made it unsuitable 
for a JN25 attack. But it reduced the number of 
required components. That made the use of the 
inexpensive and reliable simple relays practical. 13 

~t9) The search for a high-speed Selector was 
not ended, however. JN25 and the intellectual 
challenge of the Selector problem had captured 
the attention of many atOP-20-G, including 
Howard Engstrom. He gave the green light to two 
very adventurous projects. Both tried to push 
existing technologies far beyond their limits in an 
attempt to find the high-speed memory and cir- 
cuitry that an operational Full (five-digit) Selector 
would need. 14 

(U) Walls of Pipes and Tliousands of Dots 

(U) The development of radar during World 
War II had led to a very unusual memoiy device, 
the acoustic or "sonic" delay line. The delay line's 
job was to hold and recycle signals so that a radar 
operator's display screen could have refreshed 
and stable images. The "lines" were tubes filled 
with chemicals. At each end of the tube was a 
transducer. An incoming electrical signal was 
transformed into a pulse within the tube. The 


Page 169 


transducer at the end of the tube changed the 
pulse back into an electrical signal. The chemical 
medium within the tube, typically mercury, circu- 
lated while holding the data pulses. 

(U) Unfortunately, delay lines could hold only 
moderate amounts of data; they were very tem- 
peramental about the amount of heat they were 
exposed to; and much about their behavior 
remained a mystery. But they presented data at 
rates several orders faster than other media of the 


fS)- Howard Engstrom, still in search of a 
capable machine for the Japanese codes, had 
decided to take some great chances. With Enigma 
and the Fish machines under control, he deter- 
mined it was safe to assign some of his most valu- 
able men to work on a delay-line Selector. 

(SO One of the rooms at"G's" Nebraska 
Avenue center soon had a very strange appear- 
ance. A box full of electronics stood in front of a 
wall of metal tubes. The young naval engineers 
spent weeks trying to gather the electronic 
switching system, the chemical delay lines, and 
the prototype calculating units into a functioning 

(S) While the group in Washington was on its 
adventure, something more technically coura- 
geous was taking place within the secret rooms of 
the NCML in Dayton. Two of "M's" brightest engi- 
neers had been allowed to work on a veiy special 
version of the Selector when they were not busy 
with emergencies. Ralph Palmer, the engineer 
from IBM who later played a critical role in its 
computer history, led a team that was attempting 
to build a magnetic memory and advanced photo- 
optical Selector. 16 

(•6} The Palmer-Reid Selector seems a very 
strange contraption today, but in the mid-i940s 
their prototype was seen by visitors to Dayton as 
an exciting alternative, partially because it was 
another attempt to develop and apply electrostat- 

ic memory. Their Oscillograph Full Selector had 
the potential to become one of the most powerful 
and fastest of all the RAM machines. 

(S-) Their Selector was to consist of 100 mag- 
netic disks (a technology yet to be born), elec- 
tronic circuitry, a heat-sensitive printing system, 
and ten very special oscilloscopes. The disks were 
to be divided into two sets of fifty each, one set for 
possible additives and the other for message text. 
One hundred groups would be on each message 
disk. The two sets of disks were to spin in syn- 
chrony, then be offset to accomplish a full overlap 
test. Advanced electronic circuitry would differ- 
ence the two data streams and then select one of 
the ten oscilloscopes. Those ten "memories" were 
to hold the 100,000 code weight entries.* 7 

-fS)The electronic circuits of "Palmer's special 
project" would, through a coordinate system, 
select one of the 10,000 spots on the face of the 
proper oscilloscope, then turn processing over to 
an analog system. 

-4S}-The electrostatic storage was to be very 
smart. Each of the screen's dots was to have one 
of a number of possible densities representing the 
assigned weight for each code group. To register a 
score, the oscilloscope was then to be imaged 
onto a photographic mask. The amount of light 
passing through the mask would be proportional 
to the code's weight. A photocell system would 
sense the amount of light and then throw a par- 
ticular amount of current to a condenser. When 
all the groups in an overlap had been tested, the 
amount of charge on the condenser would serve 
as a measure of the probability that correct addi- 
tives had been located. 

4$}- Another subsystem in Palmer's Selector 
was to be used to dump a charge onto a "master" 
condenser and, at the same time, 200 others. 
Each of those 200 was a "memory" for the good- 
ness of each of the possible overlap tests. The 
amount of the charge on each of the condensers 
would determine how long its particular associat- 

Page 170 



ed printing head would rest on the teledotos 
paper in the Selector's printer. The greater the 
charge, the longer the print line. 

-(Sf Palmer did not complete his machine 
before he returned to IBM to lead many of its 
computer projects, including its magnetic "tape 
processing" developments. But his Selector proj- 
ect was not wasted. His experiments with mag- 
netic disks provided a basis for the navy's pivotal 
magnetic recording development projects after 
the war. 

(U) The Relay Selector Gets an Electronic 
Face L\ft 

-£ES-) While the delay line and oscilloscope 
designs were being drafted, the final design and 
construction phases of the safe-and-sure simple 
relay Selector continued. But as the machine's 
design progressed, the commitment to a pure 
relay technology or to relay switching with a resis- 
tor memory dissolved. The engineers wanted to 
experiment, and they were allowed to do so as the 
Pacific war was ending. The Selector became a 
conglomeration of old and new technologies. 
Within a few years after the war, it had grown to 
be, like Madame X, a room full of relay banks and 
plugboards, but it had a special addition, digital 
electronic components. 

•£S8}The shift to the use of some digital elec- 
tronics came as a result of an increased trust in 
the technology and a realization that a pure relay 
machine would be too slow. But the Mercury Full 
Selector of June 1945 was neither a showpiece 
electronic device nor an example of advanced 
engineering imagination. Mercury did move away 
from analog calculation, but it was relatively slow, 
and it continued in the OP-20-G tradition of 
using the least resistant technological combina- 
tions. However, it was hoped that the machine 
and the cryptanalytic method it embodied would 
justify building a fully electronic version. 18 

(TS//SI) When Mercury first appeared, it was 
a quite impressive seven-foot high by fourteen- 
foot-long bank of relays that was served by two 
cabinets of electronics. It grew even larger. 
Within a few years it approached the size of 
Madame X. Its expansion was due to postwar 
operational cryptanalytic needs and the contin- 
ued reluctance to build an electronic code 
Selector. The number of Mercury's relay banks 
was increased tenfold in order to turn it back into 
a five-digit machine and to expand its dictionary 
of stored code weights to the required 100,000 

g ES//0 f) The front end of the mid-1945 
Selector was quite conventional. An IBM collating 
machine with its two card readers was the input 
device. Code and possible additive were read in 
simultaneously, one value on each card plus an 
identification sequence. The readers were not 
exceptionally fast. In fact, descriptions of the 
machine bemoaned the Selector's slowness 
because of the low speed of the collator. 

CT0//C1 J The next part of the machine was a 
bit more innovative, but it was a mixture of the 
old and new. The pulses from the card reader 
were passed to eighty gas-filled tubes. But they 
were connected to a small relay matrix that was, 
in turn, connected to familiar plugboard matri- 
ces. They were called upon when false addition or 
subtraction was needed for additive stripping. 

(T0//0I) The number that emerged from the 
plugboard then entered a tree-like structure of 
some 1,000 relays. That was the "selector" in the 
system. The relay system then passed the code 
value to a true technology throwback, a bank of 
plugboards with 10,000 entry points. The plug- 
board banks were Mercury's "memory." 

£E S//£ff ) Each of those points was in turn con- 
nected by a plug wire to one of twenty "weight" 
relays. Eveiy time the profile of weights changed, 
the engineers had to rewire those connections. 
Each known code group had a value from zero to 


Page 171 


nineteen associated with it. Despite some tricks 
that reduced the number of code values that had 
to be plugged in t change of w 'ng < as a mas- 
sive job that took several days o effort. Pes haps 
the difficulty of the plugging was one reason why 
the 1945 Mercury was restricted to a memory of 
only ten thousand "weights." But the replugging 
was probably less ime-consuming overall than 
trying to maintain and change a condenser type 
of memory. l9 

(TS//SI) Some parts of Mercury were techno- 
logically up to date. After the weights left the 
relays, Mercury began to be something of an elec- 
tronic digital processor. Tnside one of its cabinets 
was a large electronic ring counter (something 
quite like what Bush had used in his Comparator) 
that summed the digital values that were "select- 
ed" by the suspected plain codes. Next to that 
counter was another one quite like it, but the 
"threshold" digital electronic component was 
unusual for a "G" machine. With the aid of a plug- 
board and a rotary switch, the second ~ing count- 

er could calculate a simple regression equation 
(Y - a ->- bx). The resulting value, which changed 
as each card was read, served as a benchmark for 
a test of significance of the accumulated weights. 
The parameters of the equation were usually set 
to the average weight value of accumulated mes- 
sages. 20 

CIS/ /SI) A third set of electronic tubes, called 
the "overlap counter," counted the number of 
weights sent to the accumulator during a run. 21 

( TO// 31) When the electronic evaluation unit 
that stood between the accumulator, the overlap 
counter, and the regression unit was activated (it 
could be set to check the results after every card 
was read), Mercury became a "smart" machine. 
The machine itself decided what was or was not a 
set of probable additives. If the accumulator- 
overlap balance did not match the value in the 
comparison unit, no results would be printed. 
The machine might also be ordered to automati- 

Page 172 



cally run a new set of cards that had been stacked 
in back of the first deck. 22 

t*!3//0I}- The postwar Mercury, although an 
ugly kludge, proved useful to the navy for almost 
half a decade. It was used for cipher vs cipher 
attacks and was even coaxed into becoming a ver- 
sion of the old Gee Whizzer. It could be made to 
test for the frequency of digraphs and thus give 
insights into transposition systems. 23 

(U) The Biggest Snakes of All - The Navy 
Almost Builds cm Electronic Bombe 

(U) While the "M" group at OP-20-G contin- 
ued to search for machines to breach the 
Japanese code systems in 1944 and 1945, they 
and Friedman's crew had to respond to new chal- 
lenges posed by the enciphering machines of the 
Axis powers. Germany threw the most curves at 
the cryptanalysts in Washington, but the 
Japanese also made changes in their systems that 
led to a search for new RAM. 

(U) The complex analogs of the Japanese 
cipher machines that "G" constructed during 
1943 had proved very helpful, but they were not 
analytic machines. They essentially were decryp- 
tors, machines to be used after a system had been 
solved. The cryptanalysts wanted more: a 
machine to attack the systems, especially the 
JN157 enciphering device, Jade. 

(■T0//8I) Busts and other operator errors had 

led to a general knowledge of the machine, to the 
ability to guess daily "wheel" orders and stecker 
settings, and, by late 1943, even to the discovery 
of the wiring of its stepping switches. 

( TS//SI) All that presented a tantalizing 
opportunity for analysts such as Frank Raven and 
Lieutenant Braun, but also frustration. They still 
had to find the starting positions of the important 
parts of the machine in order to read the Jade 
messages. The task was formidable. In its worst 
moments "G" thought it might have to explore as 

many as 10,000,000 to 30,000,000 possibilities 
for each daily "system" even though its attack was 
based upon cribbing. 24 

£ B)//Dr ) That demanded too much of the tab- 
ulators, even of the NC machines. So Lawrence 
Steinhardt was asked to devise a "Grenade" for 
the Japanese cipher machines, or at least one for 
the stubborn JN157. 

(U) He quickly chose a name for the proposed 
machine. He called it "Rattler." But it took some 
time before the architecture and hardware of 
Rattler were selected. There were many twists 
and turns before Rattler became an electronic 
version of a "bombe," at least a bombe for the 
Japanese stepping-switch problems. 

(TS//SI) Because of the pressure to deliver a 
machine as soon as possible, Steinhardt at first 
wanted the NCR group assigned to build Rattler 
to use standard technologies. He wrote Joe Desch 
in early 1944 recommending that Rattler was to 
be "entirely nonoptical and non-electronic in 
character." Although it was to have old-fashioned 
components, Steinhardt thought it could perform 
the required minimum of 10,000,000 tests with- 
in eleven minutes. 25 Calling on in-hand technolo- 
gy had a greater benefit. By using the electro- 
mechanical stepping switches from Viper and 
some relays and plugboards, Steinhardt thought a 
Rattler that tested a short crib against cipher 
could be in operation within three weeks. 

(U) But some disadvantages to using old com- 
ponents surfaced and the construction of Rattler 
was delayed. The drawback to the first proposal 
was that it called for the coordination of 108 
electromechanical stepping switches. Joe Desch 
thought that a bit too much too ask. He also won- 
dered if the stepping switches could be made to 
work as fast as Steinhardt imagined. After review- 
ing the first design with Desch, Steinhardt also 
had doubts. So he approved delaying the project 
while other options were explored. 


Page 173 


(U) After two very tension-filled days, 
Steinhardt presented another design. It was even 
more committed to old, trustworthy technolo- 
gy. 26 Steinhardt's second design was also driven 
by the need to deliver a machine to the eryptana- 
lysts within a few weeks. 

(U) He had thought of a handy alternative to 
the stepping switches. As a substitute for at least 
some ofthem, Steinhardt suggested that Desch 
develop what Howard Aiken had used on his 
Harvard-IBM protocomputer, a very high-speed 
tape version of the IBM card. 

(U) Using the uncut IBM paper stock as the 
input medium and six slightly modified versions 
of the readers from IBM sorting machines, it 
would be possible, Steinhardt claimed, to elimi- 
nate most of the stepping switches and perform 
the crib tests in perhaps half the time the first 
design required. The six input tapes would be 
representations of the letter developments of the 
crib letters. 

(U) Joe Desch considered the second propos- 
al and quickly responded with a long list of objec- 
tions and alternatives. Asa result, the Rattler 
became something much more technically 
advanced than anyone had imagined a few weeks 
before. But as a result of Desch's recommenda- 
tions, it took an additional half year to turn 
Rattler into an operational machine. 

(U) The Rattler that emerged was very, very 
different from Steinhardt's early conceptions. 
Rattler became one of the most advanced elec- 
tronic machines of the SIGINT war. The necessi- 
ty for speed drove Steinhardt and Desch to take 
the risk of relying on electronics. 

(U) The electronics needed for the JN157 
problem was much less demanding than for the 
Enigma, however. The critical component of 
Japan's Jade machine was a telephone stepping 
switch which had, almost, twenty-six possible 
positions. Its electronic analog needed the same 

number of "positions." An Enigma wheel was a 
much more complex mechanism to imitate. To 
mimic it called for a matrix of over 670 "posi- 
tions" and allied circuits. That meant approxi- 
mately 1,000 tubes to imitate an "E" commutator. 

CfS//3f) The limited number of tubes needed 
to imitate a stepping switch allowed Desch and 
Steinhardt to think that an electronic Jade 
"bombe" was a possibility. With faith in their abil- 
ity to build electronic "rings" because of their pre- 
vious work on the counting circuits of the 
Comparators, they began to design a minimal 
version of an electronic crib tester. Electronics 
was used where essential, but some of the oldest 
technologies were incorporated when they 
proved most efficient. 

( TS//SI) The Rattler that appeared in May 
1944 was an electronic stepping-switch version of 
a very, very fast crib testing bombe. But it was 
limited in function, as were the Enigma Bombe's 
Grenades. In addition to demanding a known 
wheel order, Rattler needed to be told what 
"stecker" had been used. It did not have a diago- 
nal board test as did the Bombes. 

(U) Rattler had at least 1,000 tubes. Its heart 
was six banks of electronic stepping switches 
(ring counters) which were analogs of the electro- 
mechanical versions used by the Japanese. There 
was another electronic component, the large cab- 
inet of detector circuitry used to identify a hit. 

fflj//GI ) A huge bank of lights allowed the 
machine's operator to see the positions of the var- 
ious stepping switches when crib matched cipher 
text. 27 Rattler had other technological throw- 
backs. The electronic switches fed into banks of 
relays, and much of the setup of the machine was 
done through rows of plugboards. Twenty-five of 
them were used for the final "switching" function 
of the two "fixed" steppers in Jade. 

(U) Rattler was a technical and operational 
success. It was able to run through all the posi- 

Page 174 



tions of the switches and test for a crib-cipher 
match in ten minutes. It proved so valuable that a 
second model was built and modifications soon 
allowed Rattler to be used to attack other 
Japanese cipher machine systems. 

(U) But Rattier did not turn out to be as small 
as Lawrence Steinhardt had promised. In fact, it 
was a giant; it was seven feet high and nine feet 
long. 28 However, it was puny when compared to 
two other machines Lawrence Steinhardt began 
to pursue during late 1944. 

(U) The Serpent and Friends 

(U) Lawrence Steinhardt had been frustrated 
throughout the war by having to continually react 
to emergencies. He became tired of hastily build- 
ing machines he saw as crypto and engineering 
compromises. Like his ex-mentor, Vannevar 

Bush, he felt that the navy should have a stock of 
a few types of fast and versatile machines that 
were suitable for the full range of cryptanalytic 

(U) In 1944 Steinhardt was able to spend 
some time on that concept; by the end of the year 
he had a proposal for a machine that would sur- 
pass the Comparator's ability to attack many dif- 
ferent types of problems. He called his all-pur- 
pose machine "Serpent." 

(U) The Serpent would have been a great sur- 
prise and a disappointment to Vannevar Bush: it 
turned against his favored technologies and 
favored reliability and flexibility over speed. 
Worse, it was to be centered on IBM components. 
Steinhardt openly declared that photoelectric 
technologies were too temperamental and micro- 


Page 175 


film too demanding, at least for a machine that 
needed many simultaneous inputs. 29 

fS} Serpent was to have at least thirty-two 
input stations. Each would read the "tapes" made 
of uncut IBM tabulator card-stock formed into an 
"endless" IBM card. The reading heads were to 
each have eighty brushes so that each hole in the 
"tape" could be read simultaneously. In each of 
the reading stations, as many as 100 of the heads 
could be installed,. How many were to be active at 
one time was to depend upon the cryptanalytic 

($Hn addition to the thiity or more loo-level 
reading stations, at least two more would be 
available for multitape operations such as done 
on the Comparator. 30 

(S-) A control system was to allow the tapes to 
be driven synchronously, or in any of the 
Comparator motions (stepping-sliding), or in 
such a manner as to imitate a matrix. The varied 
stepping would allow Serpent to have many uses. 
It could be a Copperhead or a Comparator or a 
Bombe or a Rattler or an IC machine - or even a 

(U) Steinhardt's Serpent rejected more than 
photoelectric reading technology and microfilm. 
He did not want to bother with electronic count- 
ing, either analog or digital. Because the 
machine's input was relatively slow, the rate of an 
IBM sorter, there was little need to bother with 
the pesky tubes. Rather, Steinhardt recommend- 
ed that a set of relay boxes be constructed. Each 
would perform, like the SIS IBM machines, a par- 
ticular set of functions. 

(U) Steinhardt admitted that Serpent would 
not be able to perform some attacks as fast as the 
advanced photoelectric RAM, but its chameleon- 
like quality would, he claimed, more than com- 
pensate. It would, he said, be a perfect type of 
machine for research and for the postwar era 
when emergencies no longer drove OP-20-G. 

{&) In some cases, he said, it could compete as 
an operational machine. While Serpent would 
take three or four hours to do a full four-wheel 
Enigma run (compared to twenty minutes on the 
Bombes), it would, according to Steinhardt's cal- 
culations, be as fast as the electronic Rattler on 
the Jade and Coral problems. As well, he said, 
Serpent would be as fast as most of the photo- 
optical machines, at least the ones that used 
punch tape or photoplates. When the time need- 
ed for photoprocessing was taken into account, 
Steinhardt claimed, Serpent would be as efficient 
as the microfilm Tessie and Hypo. 

( ■ S/Z ' Sf ) 11 But competing with those two 
machines was not important to Steinhardt; he 
and others had concluded something that would 
have offended Stanford Hooper: "The crypto- 
graphic value of polygraphs and f.C. runs . . .is 
now admittedly open to question." 31 The signifi- 
cance of Serpent would belts ability to quickly 
test out such cryptanalytic applications to see if 
they were worthwhile. Serpent would prevent 
investing in costly special-purpose machines, 
ones that had little payoff. 

(S//SI) Steinhardt concluded his report on 
the proposed Serpent with some very prophetic 
advice: Serpent would be needed for the navy's 
next great challenge, the Russian code and cipher 
systems. 311 

(U) Lawrence Steinhardt's suggestion for the 
IBM Serpent was not followed through, although 
he continued to work on it and the design for the 
electromechanical counting machine that became 
the postwar monster, Alcatraz. Importantly, he 
was returning to the fold of the believers in elec- 
tronics; he had begun work on an all-electronic 
ciphering machine, just as his colleagues were 
again forced to try to overcome the weaknesses of 
digital electronic components. 33 

Page 176 

top cr:oncTOGQM i NT;/ncL to uga, aug, gam, gor and hzu 


(U) The Revenge of the Enigma - or 
Electronics Is Inescapable 

(U) Although OP-20-G and the SIS turned to 
the Pacific after 1943, the Enigma problem 
returned to plague them. Actual and feared 
changes to the Allies' old nemesis were what 
drove the army and navy to commit massive 
resources to solving the problems of large-scale 
electronic systems. 

(T0//0I) The alterations to the Enigmas and 
their operational systems, especially the 
Luftwaffe's decision to make its reflector's wiring 
"pluggable," demanded so many tests that only 
electronics could perform the attack. The "reflec- 
tor" problem of 1945 forced the development of 
devices that came close to being electronic 

£ES#8f) At first it appeared that even the best 
technology could not overcome the new Enigma 
threat. Fortunately, a cryptanalytic attack on the 
changeable reflector was created that did not 
demand a fully electronic version of the Bombe; 
that would have been an impossible goal for the 
army, the navy, or the British. But the Duenna, 
the Superscritcher, and the Giant machines they 
constructed for the problem were "the" electronic 
cryptanalytic devices of World War II. They went 
far beyond the Comparators or even Rattler. 34 

(• TO//3 f) But the electronic solution was a 
long time coming. The Americans did not leap 
from Joe Desch's electromechanical Bombes and 
Madame X to electronic machines. They tried to 
conquer "E" operational changes and then the 
"reflector" problem with traditional technologies. 
OP-20-G made several alterations to the original 
Bombe design before it accepted the necessity of 
the electronic Duenna, and the SIS built a huge 
new relay machine before it started building its 
electronic Scritcher. 35 

(U) OP-20-G's changes to the Bombes were 
evolutionary. The first major ones came after the 

British made an emergency request for an addi- 
tional set of American Bombes, at least fifty of 
them. Joe Desch took the request as an opportu- 
nity to improve the standard #530 Bombes. He 
produced some two dozen of the new # 1530s in 
1944. They used :he same logic and technology as 
the 1943 machines, and they ran at the same 
speed as the #530s, but were mechanically 
stronger and had additional circuitry to eliminate 
false stops. 36 

(TS//SI) However, even before Desch made 
those significant technical improvements in the 
original Bombes, he began constructing the "Fire 
Engines." Those eight machines were the same as 
the original Bombes except that the vertical order 
ofthe commutators was "inverted." The fast 
wheel on the Enigma became the slow wheel on 
the Fire Engine, and the slow wheel on the 
Enigma was in the fast position in the Bombe. 
Nothing else was significantly different from the 
#530s. But the "inversion" was powerful. It 
allowed quicker runs when the identity ofthe fast 
wheel was known, and, more importantly, it 
allowed what were called "hoppities" runs during 
which the operators could stop the machine, then 
advance a wheel one step by hand. 

( ■ TG//DP ) That cumbersome process was nec- 
essary because the Bombes were unable to auto- 
matically imitate the turnover action ofthe "E" 
wheels. When Enigma wheels reached a certain 
position, they "kicked" the adjoining wheel one or 
more steps ahead, thus breaking the regular met- 
ric motion ofthe Enigma. 37 The Fire Engine 
"hoppity" method was very crude, but very help- 
ful. It allowed the use of weak menus and ones 
which travelled over probable turnover positions. 

(TS//SI ) A more complex extension ofthe 
commutator Bombe was Grandad, the double 
unit Bombe. It had thirty-two, not just sixteen, 
"E"s linked together. The use of twice as many "E" 
units in Grandad decreased the probability that 
an incorrect setting would result in a "hit." 
Arriving in Washington in late 1944, Grandad 


Page 177 


permitted the use of much weaker cribs than 
demanded by either the regular or the inverted 
Bombes. It was designed to find solutions when a 
set of short indicators was used as the crib, when 
there were unknown stecker connections, or 
when the crib consisted only of cipher letters that 
were known to represent the same plaintext let- 

(U) Beyond Cribs; the Statistical Bombe 

(U) The most ambitious revamping ofthe 
Desch Bombe was the Bulldozer. Delivered in 
early 1945, it had been desired, if not planned, 
since the navy first accepted the commutator 
Bombe in late 1942. It was a mechanical answer, 
and a very clever one, to the demands that *G" 
move towards a pure attack on "E," 

(■Sf/Sf) Although everyone at OP-20-G had to 
accept using a crib-based method against 
Enigma, many argued for a continued search for 
a "pure" attack. Some were committed to statisti- 
cal analysis out of professional pride; others cau- 
tioned against the danger of depending upon 
Britain for cribs. They warned that if the Germans 
tightened their security, even GC&CS could not 
supply what the standard Bombes needed. 39 

£jSjf$T) In early 1943, a search for what was 
generically called a "statistical" solution was 
begun, but it was a very limited effort. 4 " There 
was too much else to do to allow anyone within 
"G" to focus on an abstract problem. But when 
there were hints in summer 1944 that the Allies 
might not be able to count on gcod cribs in the 
future, more resources were poured into develop- 
ing a machine for something very radical: a 
cipher-only attack. 41 

(TS//6F) B ecause at least a prototype machine 

was desired as soon as possible, Joe Desch's crew 
was asked to see if a regular Bombe could be 
turned into a "statistical machine." It took some 
time to refine the method and to revamp the 
Bombes, but a Bombe to identify German plain 

language became operational in March 1945. It 
was called Bulldozer because ofthe mechanical 
power a cipher-only attack demanded. 

■ £TS//£r F) The impressive Bulldozer was a 
cross between the hardware of a double Bombe 
(Grandad), the logic of a Mercury Selector, and 
the electronic analog circuits of an IC machine. 
Although it used a somewhat less mechanically 
demanding "recognition" method than did 
Mercury, Bulldozer's power to identify probable 
plain text was significant. 

(3S//ST) Much energy had gone into devising 
the algorithms that had to be built into a crib-free 
Bombe. Hundreds of hours were put into the 
analysis of German military language. And "G's" 
best mathematicians spent weeks integrating 
those findings with probability studies to arrive at 
Bulldozer's test for the appearance of plain text. 42 

gF&//0I ) The method finally embodied in 
Bulldozer demanded much that was new. 
Bulldozer's test began with the entry ofthe inter- 
cepted cipher. Then the machine's wheels were 
spun. At each position, the letters that emerged 
from the wheels were electrically weighted and 
summed. Next, a comparison was made as to 
whether the square ofthe weighted frequencies of 
each letter summed to equal or exceed a value 
that was typical of good plaintext messages. 43 

6R3//3fl rThe earlier statistical studies had 
determined that Bulldozer needed along crib to 
be able to differentiate random text from true 
"language." Thus, the machine was, like Grandad, 
a double Bombe. To give the new machine addi- 
tional power to tell order from chance, its thirty- 
two double banks of four-wheel "E" units could be 
changed into sixty-four single units to accommo- 
date a longer cipher. Bulldozer was also like the 
Fire Engine: its banks were inverted. That was to 
help make "hoppity" type runs. Significantly, 
unlike the other Bombes, Bulldozer did not have 
a diagonal board." 

Page 178 


\W SECRnWCOWttflWJCt J »W«^-A«9r9iV4^»»AN s ?4aU*U 

<iD//ar ) 

to iiiiiLify 


4TQ//3T) Bulldozer called on electronics as 
well as electromechanics. After the cipher had 
been set on the machine's dials, all the letters that 
emerged from the commutators were sent to a 
small bank of twenty-six tubes. They stored the 
accumulated electrical weight for each letter. 
Before any of the wheels were moved, the value in 
each tube was squared and passed to an analog 
summing circuit. When the combined value of the 
output from the commutators at a particular set- 
ting exceeded the assigned threshold value, the 
machine stopped and then did something quite 
different from the other Bombes: it printed out 
the full text of the deciphered crib on a 
Letterwriter typewriter. 

(TG//0I) - Bulldozer's pure attack took more 
time than a crib vs cipher one. At its very best, it 
took twice as long to run a grenade test. That was 
because its motor was set at one-half the typical 
speed of the Bombes. That limitation was com- 
pounded by the nature of Bulldozer's tests. When 
the frequency weighting system confronted an 

uncooperative cipher, the machine might stop 
and type out probable clear text so frequently that 
its running time increased to as much as eight 
times that of the regular Bombes. That was one 
reason it was rarely used for more than a grenade 
run to establish starting points after all the other 
Enigma settings had been discovered. 

G£S//Sf) Full Bombe runs were much more 
forbidding. Given the special assumptions that 
had to be made about the stecker in a full run, 
Bulldozer might have to make as many as twenty- 
six separate four-hour runs to produce a solu- 
tion. 45 

(TS//0I) Although it probably never broke an 
Enigma system, the cryptanalysts were quite 
impressed with Bulldozer. Its weighted test 
seemed so promising for the future that "G's" 
cryptanalysts informed OP-20 that a Bulldozer 
attack might well make the navy's own advanced 
cipher machine, the ECM, vulnerable. Bulldozer 


Page 179 


seemed so able that "G" did not tell the British 
about its powers until well after the war. 46 

(U) No Escaping Electronics, Enigma Meets 
the Cobra 

(U) It was not a fear of what the Germans 
might possibly do with Enigma in the future but a 
true emergency that finally drove the army and 
navy's cryptoservices to take a chance on large- 
scale electronic machines. The emergency arose 
in 1944. It was the German air force's change in 
the internal workings of its three-wheel Enigma. 

(TS//SI) Fortunately, the Germans had given 
the British codemen some hints in late 1943 that 
the Luftwaffe was going to switch to the use of a 
pluggable reflector. That had caused a great deal 
ofworry. Somewhat later, when the British 
learned the German Army was to do the same, 
worry turned into near panic. 

(T3//3I) 1 The new reversing wheel was a 
major threat. With the pluggable reflector, any 
letter could be quickly rewired to produce any 
other letter. Although the new reflector was sta- 
tionary once it was placed within an Enigma, its 
ability to be any possible wheel made it worse 
than the fourth wheel in the M4. It called for 
examining an additional 150,000,000,000,000 
possibilities when attacking the air force's and 
army's "three wheel" machines. 

ffig//S 9 The challenge was daunting, and 

there were some thoughts of not even attempting 
to conquer the pluggable wheel. But when it was 
learned that the Germans would alter the wheel's 
wiring on a ten-day cycle, not every day or with 
every message, a decision was made to face the 
problem. 47 

(■EfyfSQ The British were the first to attack 
the new wheel. They did the best they could 
against the German communications subsystems 
which employed the new reflector, but the chal- 
lenge was too much for their limited resources. 

Under the best conditions, it took five top 
Bletchley Park mathematicians two weeks of 
"hibernation" to get a solution to a single 
rewiring. The most powerful of the British 
Bombes were of little help. A slightly modified 
one took sixteen days for a simple problem, and 
the more elaborate Giant took three to four weeks 
to complete a full menu. That meant that 
Luftwaffe traffic could no longer provide much 
important tactical information. That traffic had 
become vital to the Allies. With the end of the U- 
boat war and the retreat of the German army into 
the homeland where messages were carried by 
cables, air force transmissions were a prime 
source of radio intelligence. 48 

(T3//3I) A technological solution was 
required. But in mid- 1944 the British were tired 
and stripped of resources; they needed help. 
Unlike the situation in 1942, they did not hesitate 
to inform their American cousins of the new dan- 
ger and of possible methods of solution. The navy 
was notified of the "scritcher" test, then the army. 
Both American agencies immediately got to work 
trying to turn the probability-based scritcher 
method into hardware. 

(U) The Navy's Duenna 

( TS//SJJ The navy's first thought was to make 
another simple modification of the commutator 
Bombes. It was to be called "Mona." Mona was to 
have just one traditional Bombe wheel. The rest 
of the Bombe was to test the huge number of 
assumptions needed to try to identify the reflector 
wiring and the stecker pluggings. Mona quickly 
proved itself incompetent. 50 

(3 S//SI) More thought went into the method 
and possible machines. After modifying one of 
the Enigma analogs at "G" by adding the "Cobra," 
an attachment that allowed the analysts to change 
the reflector wrings by hand, "G's" analysts put 
the British method through a series of explorato- 
ry tests. Soon, they decided that only a two-wheel 
test and a great deal of electronics could do the 

Page 180 


tTTsscxspownr mt mmA, au&, can, «m mtmtm 


(TO//DI ) Duama 

job. Preliminary specifications were drawn in 
August. Then, very quickly, the proposed 
"Duenna" was under construction at the NCR 

(TS//3I) Duenna became a twenty-foot-long 
and eight-foot-high mixture of a small version of 
a two-wheel commutator Bombe and a cabinet 
with over 3,000 advanced electronic tubes. The 
first Duenna of November 1944 had to be huge 
and cutting-edge because of the demands of the 
scritcher test; so did its four sisters. 51 

(TS//SI) One of "G's" bright young mathe- 
maticians, Howard Campaigne, made important 
contributions to the logic of scotching. 52 His 
modified OP-20-G 53 version of the British attack 
was hardwired into Duenna's complex circuits. It 
was necessarily a very efficient attack, but it 
demanded a great deal. It required as much as a 
100-letter menu (cribs vs cipher pairings in this 

case); it worked under the assumption that the 
slow "E" wheel would not turn over at critical 
points in the menu; and, unless prior knowledge 
allowed "G" to avoid it, a full test required as 
many as fifty-six separate runs to test all the pos- 
sible two-wheel combinations. 

CE S//0 JJ Duenna demanded so much because 
it was expected to work much harder than the 
original Bombes. The Bombes were pampered. 
They were well fed with known wheel wirings and 
the reflector plugging and asked to yield stecker, 
rotor order, and, with some help, the window set- 
ting. All that Duenna was told was the wheel 
wirings. Yet it was asked to produce the same 
information as a successful Bombe run plus the 
plugging pattern of the new reflector. 

(TS//S J- ) Duenna's job was tough, and it had 
to be "smart." With the crib in place, it made an 
assumption about the stecker setting of two or 

top SEen c T/fe o M i N T rm e L to usa, aus, can, gb r and nzuia i 

Page 181 


three high-frequency letters. Then, with its two 
imitations of the faster "E" wheels, its electronics 
serially tested the plain-crib pairs against succes- 
sive combinations of stecker and reflector plug- 
gings. Duenna had several electronic versions of 
steckers and reflectors to perform those tests. 54 

(TS//GI) A stecker-reflecter check for a given 
stecker assumption took twenty minutes; an 
average "run" took an hour and a half. But the 
need to run different wheel combinations to iso- 
late the fast wheel turnovers called for a day's 
work. The complete set of tests needed for diffi- 
cult systems might keep a Duenna busy for two 
weeks. 55 

"links and tacks" test was passed, Duenna printed 
the information about its "hit" and went onto 
search through another stecker assumption. If 
the threshold was not met, Duenna skipped the 
printing stage and went on to examine another 
stecker assumption. 56 

( T I 3// 31 ) Although it was a very strange look- 
ing combination, Duenna became a valuable tool. 
Even before the first model was completed, the 
navy ordered that several clones of the original be 
delivered as soon as possible. The crew at Dayton 
wished to build a more elegant machine, perhaps 
with electronic wheels, but they followed 
orders. 57 

(JS/r'Sf) Although Duenna had advanced 
electronics and circuitry, it might have taken 
much longer to do its job if the "scritcher" test had 
not been so inherently intelligent. Other of'G's" 
World War IT machines had hardwired decision 
algorithms, but Duenna had the most complex 
one. Duenna's smartness was based on some- 
thing like the diagonal board test: it looked for 
contradictions or impossibilities given the data 
and assumptions fed into it. Its intelligence went 
beyond the diagonal board's, however. Because 
the searching in Duenna followed a logical pro- 
gression, a "branch" of a search could be aban- 
doned very early. The machine did not have to 
wait until all possibilities had been examined. 
That saved enormous amounts of time. 

HTt^/SI) As the machine tested the assumed 
and, then, the recovered steckers and reflector 
links against the crib-plain pairs, it checked for 
contradictions as well as building up a "recovery 
number." The recovery number was changed with 
each sweep (step) through the "E" analog, That 
number was compared with the recovery score 
from the pre 1 ions sweep. If no additional recov- 
eries of compatible steckers and reflector plug- 
gings had been made, Duenna's circuits com- 
pared the ic.ccvery "buildup" number with a pre- 
set threshold value. Tfthat value was exceeded, 
Duenna tested for another critical factor. Ifthat 

(S//SI ) But there were attempts to persuade 
"G" to allow the creation of a full electronic 
Duenna. 58 Some looked upon the Duenna project 
as the challenge that would force a do-or-die com- 
mitment to such electronic components. But an 
electronic "wheel" again proved too much of a 
challenge for its "E" wheels. 

( TS//SI) What was inside Duenna's main cab- 
inet was very different. The vacuum and gas-filled 
tube circuits were innovative. There were high- 
speed digital counting rings, "selection" matrices, 
electronic stepping switches, and an electronic 
version of the conflict-testing diagonal board sys- 
tem. Duenna's critical threshold testing circuit 
was very advanced for the era, and the ability to 
enter the crib via dial switches saved much set-up 
time. 59 Especially important, the Duennas proved 
very ™eliable and maintenance free. 

(U) From Relays to Tubes, Rosen Gets His 

CFS//S-B The SIS engineers under Leo Rosen 
also found the pluggable reflector problem an 
exciting challenge. In fact, it was the difficulty of 
the problem that led the "F" section to become a 
significant force in research and development. 
The pressure on the STS to deliver a machine to 
help the British, combined with the decision not 

Page 182 



to hire an outside contractor as had been done 
with Madame X, gave Rosen the power to require 
the army to send him a cadre of bright young 
engineers who already had experience in 
advanced electrical and electronic engineering. 
He very quickly built up a remarkable group, 
many of whom went onto become important fig- 
ures in the postwar computer industry. 60 

(TS//SI ) Rosen had been notified of the plug- 
gable reflector (Uncle Dick) threat in early 1944. 
While the SIS cryptanalysts were put to work 
scratching by hand methods, he and his close 
aides, such as Captain C. R Deeter, 61 began a 
design for a supermachine. Like the navy's engi- 
neers, Rosen's men thought electronics was the 
logical route. 62 Then they were forced to weigh 
engineering ambitions against cryptanalytic 
needs. The result was another compromise - the 
relay Autoscritcher (also known as Grapevine) of 
1944. 63 And even its follow-on, the electronic 
Superscritcher, had to face up to the limits of dig- 
ital electronics in the mid-i940s. 

OS//St) Because the German army and air 
force "E" modifications were the same and 
because Britain had provided both the American 
army and navy with the same crypto-methods for 
an attack, the logic and architecture of the SIS's 
device were similar to Duenna's. But the hard- 
ware was very different: a longer crib was used, 
the build-up to a hit was different, and the test for 
branching out of a search was simpler. 64 In sever- 
al ways, the army's relay Autoscritcher was more 
of a throwback than Duenna, and it was slow: 
"running time was as much as ten to fourteen 
days, three shifts a day." Although it performed 
twenty-five tests a second, a single wheel order 
test took three and one half hours. 65 

£ ES//Di j The most incongruous part of the 
Autoscritcher was its "wheel" unit which fulfilled 
the same function as the rack of commutators in 
the Duenna. The army machine had twenty banks 
oftwo special Enigma wheels, not commuta- 

tors. 6 Each served as the moveable fast wheel 
and as a static combination of the medium, slow, 
and reflector wheels. These rotors were grouped 
into sets which, along with allied relay circuits 
and stepping switches, formed the basis for 
"branch" testing. 

CES//ST) To save design and construction 
time, no automatic stepping controls for those 
rotors were included in the Autoscritcher. Given 
the length of time it took the machine to search 
through steckers and to test for contradictions, it 
was thought 67 hand-turning of the rotors would 
be acceptable. 

( TS/ /8f) The Autoscritcher's very long crib 
was entered on a huge plugboard. Assumptions 
based on frequent cipher-plain pairs were 
entered into the machine. Then a bay of relays 
and electromechanical stepping switches serially 
tried all the possible steckers, passing control to 
the two bays of electronic equipment. They 
included the matrices that checked for diagonal- 
board type contradictions. Along chain of crib- 
plain pairs was used (each with its own rotor 
pair). An analog threshold test made the final 
decision as to whether a "hit" had occurred. That 
test, however, was not as complex as in the 
Duenna. And an operator had to copy the "hit" 
settings by hand. 68 

(U) Engineering l*ride and Peacetime 

(3//3I ) The SIS engineers had begun the 
Autoscritcher as a learning project. Once they 
tested out their ideas with its older technology, 
they expected to begin the design and construc- 
tion of an electronic machine that would be more 
useful and which would be a source of engineer- 
ing pride. The goal was a purely electronic 
machine that would be at least one hundred times 
faster than the Autoscritcher." 6< * 

CES//SH The "F" group was given the 
go-ahead for such a machine in early 1945. 

top GEonnv/oo Mi mv/ncL to uga, auc, oa n , g b r and nzu/xi 

Page 183 


The machine soon acquired the name 
"Superscritcher." Its designers were thrilled when 
they realized that it might be as much as 500 
times faster than its relay predecessor. 70 

( S//8 fr) Designing new circuits was exciting 
for the young engineers assigned to the "Super" 
project. Although the "Super" had the same archi- 
tecture as the Autoscritcher, made similar hard- 
wired "if decisions, depended upon the plastic 
and copper "E" wheels, and performed the same 
type of cryptanalytic test, it was a far different and 
more innovative fixed-purpose computer. 71 A 
purely electronic stepping switch, improved elec- 
tronic ring counters and circuits ^ and electronic 
controls for automatic stepping of the rotors 
emerged before VE Day. 

Cf &ff Q ifl The electronics in the "Super" were 
not exclusively digital, but they were major 
advances in the state of the art. Over 3,500 tubes 
were mounted in eighteen eight-foot-high bays. A 
large air conditioning system stood by to protect 
them and the operators. There even was a means 
for printing the settings when a "hit" was found. 73 

( j TOy/g-f) But the "Super" was not operable 
when Germany surrendered, and the Enigma 
traffic for which it was designed disappeared. 
There were demands that the project be aban- 
doned. Fortunately for the "F" group's morale, 
the commander of Arlington Hall Station decided 
that some useful function could be found for a 
completed machine. He approved a continuation 
of the prcjec!. Several stubborn electronic prob- 
lems were overcome and the Superscritcher's 
power was turned on in December 1945. To the 
surprise of many, it was used on various problems 
for the next five years. And it did prove itself to be 
500 times faster than the Autoscritcher. 74 

(U) Keeping the Faith; the Return of the 
Film Machines 

(U) The term "RAM" h?d a more precise 
meaning for some in OP-20-G and the SIS than 

"high-speed cryptanalytic machines." Especially 
for the army's engineers, RAM meant the type of 
microfilm and photoelectric machines that Bush 
had promised to Admiral Hooper in the 1930s, a 
series of machines the SIS group came to want for 
themselves by 1944. In fact, in 1944 it was the 
army's cryptanalysts, not the navy's, who became 
the strongest advocates for film-based devices. By 
then, the navy's men in "G" had become a bit wary 
of both film-related technologies and the difficul- 
ties involved in producing reliable microfilm. The 
navy had not abandoned Bush's ideas, but it 
pulled back from Eastman Kodak and microfilm 
in 1943. After Icky arrived, the navy did not order 
any additional film machines, at least during the 
war. 75 One reason for that was the last one on its 
1942-43 shopping list refused to be turned into 
hardware in a timely way. 

£¥8} What became known as "Amber" did not 
arrive until late 1945. It took so long because of 
the cryptanalytic difficulty of the system Amber 
was asked to attack, as well as the stubbornness of 
microfilm technologies. 

(V) The Revenge of the Codes, Again 

■ 4TS//&B ' The Japanese JN25. additive code 
had led to some desperation-driven technical 
solutions at"G." The many Copperhead I's 
searched through thousands of messages in the 
hopes of finding two identical cipher groups 
spaced equally apart in two messages so that the 
analysts might be able to identify "depths" and go 
on to break into a system. The huge and expen- 
sive Mercury, with its walls of relays, was also an 
example of trying to do the impossible. 
Thousands upon thousands of additives were 
stripped every hour; then its huge memory was 
searched for frequent code groups with only a 
glimmer of hope that a hint of a possible solution 
might emerge. 

(TS//6I ) JN25 was not the only difficult and 
demanding Japanese additive code system, how- 
ever. By the last year of the war, the attempt to 

Page 184 



read the JN37 weather code continuously raised 
more cryptanalytic and technological frustrations 
than had the fleet system. 

(TS//SI) Reading the combined weather 
report grew more and more important to the 
American navy as it raced towards Japan. The 
fleet was moving so fast that American weather 
stations could not be established quickly enough. 
The only good source of critical weather informa- 
tion about many areas was the Japanese reports. 
How valuable they were became tragically evident 
when one of Admiral Halsey's task forces was 
caught by a typhoon in December 1944. The 
storm took almost 800 American lives." ?6 If all 
the Japanese reports had been read in a timely 
way, the disaster might have been avoided. 

(TS//SI) JN37 seemed to be the most efficient 
way to tap the reports. It was the system the 
Japanese used to bring together local weather 
bulletins, then transmit them as a group. The sys- 
tem was cryptanalytically strong, but it had a few 
potentially exploitable weaknesses. The underly- 
ing four-digit numeric code for the system 
remained fairly stable throughout the war, and it 
was known that the messages were very stereo- 
typed. Cribs and code-meaning identification 
were possibilities because temperature, humidity, 
and wind speed remained much the same in var- 
ious areas during a season. That meant a great 
deal of repetition of code groups. 

(T0//0I) Some JN37 variants had been read. 
But there was a tough part to JN37, its superen- 
cryption. There were many additives, and they 
were frequently changed. With those changes 
came blackouts, such as those of 1944. In the 
JN37 version that led to so much pain in that 
year, the additive book had 900,000 entries. 77 

(TS//6I) "37" had yielded, at times, to tradi- 
tional tabulator methods supported by busts, 
knowledge of indicators, and captured data. As 
long as there were busts and captures, the old 
methods worked. But even before the traumas of 

1944, there were serious concerns that more fre- 
quent changes of the additive books and alter- 
ations of indicators might cause a permanent 

(TS//SI) In response, in early 1944 Howard 
Campaigne, one of OP-20-G's bright young math- 
ematicians, began examining the practicality of 
various pure attacks against JN37. He sought an 
attack that did not depend upon knowing indica- 
tors, the additive book, or usual cribs. He experi- 
enced much disappointment. For example, he 
calculated that an Index of Coincidence assault 
with existing equipment would take three years at 
three shifts a day for any type of breakthrough. 
Quite logically, he suggested a search for better 
methods and high-speed machines. There was 
not much progress. 78 

CES //3F ) Then the cryptanalysts had some 
good fortune. The discover)' of an active JN37 
additive book gave some hope that a practical 
method could be devised. The capture of the addi- 
tive book allowed "G's" best young mathemati- 
cians and analysts to understand the logic of the 
system used in 1944 and, among other things, 
what the probabilities were that particular code 
digits would appear in certain positions of mes- 


(ffiY/GI) A quite elaborate theory was devel- 
oped, one based upon advanced Bayesian statisti- 
cal methods. The statistician argued that if "G" 
was willing to make a very costly investment in 
calculating the "centiban" weights and creating a 
machine to apply them, then at least a minimal 
but consistent entry into JN37 could be expected. 
But the method seemed forbidding. It demanded 
too much calculation. 

(TS//SI) Other, perhaps less cumbersome, 
methods of attack were explored, including the 
dictionary lookup approach of Mercury. But they 
stood little chance of producing results. By late 
1944 the Bayesian "statistical crib" method of 

top occri:tooom in t//rcl to uqa, auo, OAN, ODR AND NZU/X1 

Page 185 


identifying a code group version of "plain text" 
became the navy's only hope. 

(TS//SI) The "statistical crib" method was the 
kind that Hooper and Wenger wanted as the basis 
for all of "G's" work. It would use only cipher and 
a "scientific" weighting system to pinpoint proba- 
ble plain text and then point to additive "depth." 
From there, additional machine runs and statisti- 
cal and craftsmen's tests could be applied to re- 
create the additive book and to quickly enter a 

i •■ So 


( TS//SD But in 1944 there was no machine 
for the "statistical crib." Although "G" had 
attempted to establish a machine program to 
make the statistical methods practical earlier in 
the war, little had emerged by the time the JN37 
weather code became a priority at "G." The lack of 
results was one of the reasons why Wenger, 
Engstrom, and Meader wanted their own crypt- 
analytic machine "factory." They were frustrated 
by the slow pace at Eastman Kodak. 

CR3//9I ) In early 1943 Eastman had been 
asked to explore possible technologies for the 
additive problems. "G" had requested film-based 
machines for a range of functions. They wanted 
ones to strip additives, match plain text to a code 
dictionary, test for likelihood, perform statistical 
weighting, and to print code meanings. 81 

(T8//SI) Eastman took up the challenge. Ten 
of the best men in Rochester were assigned to its 
new RAM team. They were given a great many 
resources and much engineering latitude. They 
explored many, perhaps too many, logical and 
technological alternatives. 

(TO//OI) But no new film "Japanese" 
machines emerged from Eastman in 1943 or 

(U) Meanwhile, Lawrence Steinhardt began 
work on his versions of additive photoelectric 
RAMs, calling upon the technical expertise at 

Dayton when he needed practical advice. By the 
time he turned his first designs into hardware, he 
had become shy of film. His proposed punch-tape 
Copperhead series was the outcome. He laid out 
ideas for a wide range of machines, ones to 
implement the newest code-system attacks. 

( T8//3I) Two of the proposed Copperheads, 
Mark V and VI, for example, were aimed at mech- 
anizing weight and dictionary attacks against 
additive systems. Projects for them were begun, 
but their complexity led to their abandonment. 
The other designs also proved too complex. Only 
Copperhead I, the "brute force" matching 
machine, was made operational. As a result, "G" 
was left without much in the way of additive RAM 
machinery. It was not alone. The army also had 
ideas of a film machine for the additive problem, 
but it gave such a machine low priority before 
1945. 82 

( T9//3I) Thus, when reading JN37 became an 
imperative for the navy, there was no hardware in 
place that seemed able to perform any of the pro- 
posed cryptanalytic attacks, especially the one 
that seemed best, the sophisticated but costly 
"statistical crib" method. 

t9//D¥ ) There was another hurried search for 
a "37" machine in 1944. None of the suggestions 
seemed reasonable, and none were turned into 
projects. The situation seemed hopeless. But at 
the end of the year, it was declared that an answer 
had to be found. 

(U) More Numbers Than Ever Before 

( ■ S/Z -SJ ^ A crisis team was put together to try to 
force a technical solution. John Howard led the 
navy group that consulted with the top engineers 
at Eastman and NCR. He told them of the 
requirements of the new "statistical crib" attack. 
Both groups then reviewed the technological pos- 
sibilities. 83 

Page 186 



(TQ//0I) What Howard gave as the goals for 
the new RAM was staggering. But that was 
inescapable. The Japanese were changing the 
JN37 additive book three times a year. To recov- 
er the new books, Howard explained, called for 
more calculations than "G" had ever attempted. 
The pure "Monographic Statistical Method" 
required, for example, 500,000,000 very com- 
plex comparisons of cipher to "crib" to find one 
correct line-up (the path to one correct additive). 
To recover just 10 percent of the "37's" additive 
book within the first half of its four-month life 
called for six such searches a day. 

Q¥&ffST) The job was more difficult than just 
looking for raw coincidences as the original Bush 
machine had been built to do. The "Ideal" 
machine John Howard wanted had to perform 
more than 200 multiplications, summations, and 
threshold tests for each of the half billion com- 
parisons made during a run. 84 To complete those 
trillions of operations six times a day called for 
something beyond the Bombe, Mercury, or even 
Duenna. Howard was calling for a supermachine. 

(3 S//SQ Even if a technological answer was 
found, building the superadditive machine would 
be a gamble. All the calculation might be for noth- 
ing. The eight weeks of 3,000,000,000 daily 
comparisons would yield a useful bit of informa- 
tion only if all elements of JN37 besides the addi- 
tives remained stable. If the Japanese made 
changes besides issuing anew additive book 
every four months, the calculations might prove 
useless. 85 

(U) The gamble had to be made. The "G" engi- 
neers, NCR's men, and the Eastman group were 
ordered to look at various technical options and 
come up with a solution. 

(TS//SI) One alternative was to base a 
machine on advanced digital electronic count- 
ing/multiplying circuits. When estimates were 
made ofthe speed of the best possible vacuum 
tube machine, the results were shocking. It was 

found that just one run with the electronic digital 
device would take sixteen years. When asked if 
several copies ofthe device could do the job in a 
reasonable time, the engineers responded that it 
would be impossible just to find the parts needed 
for enough ofthe machines. 86 

( TO//DI) Another possibility was to create a 
new version ofthe proposed Full-Selector 
machine. Instead ofthe weights being calculated 
as the cipher was scanned, they would be stored 
in a fast memory and retrieved for each ofthe 
possible 60,000 different cipher combinations 
that were expected to be encountered during a 
run. The stored-weight alternative seemed attrac- 
tive, but it called for ultrafast memory. Without it, 
the stored-weight approach would take as long as 
the on-the-fly calculations using electronic digital 
circuits: years. 

( TS//SI) The "best" memory technologies, 
such as delay-lines, seemed inadequate. So it was 
decided to go back several steps. The only way to 
achieve the needed speed, it seemed, was to rely 
upon very densely packed high-speed tapes that 
carried all the "weighting information" and upon 
analog and parallel calculations. With the tapes, 
the machine would not have to multiply; it would 
just have to scan, search the memory, and sum. 
And, if enough precision could be obtained, speed 
could be vastly increased through the use of ana- 
log "counting" - the machine would not have to 
wait while digits were summed, one after the 

(TS//ST) All that seemed to dictate a return to 
microfilm and photoelectric sensing and calcula- 
tion. But it took many months to agree upon the 
exact nature ofthe machine. NCR and Eastman's 
team agreed that the general microfilm RAM 
approach should be followed, but there were sig- 
nificant differences on many details. An impor- 
tant one was whether the "weights" were to be 
represented by different size spots (the NCR rec- 
ommendation) or by degrees of opaqueness. 


Page 187 


(S//SI) The debate went on far too long. It 
was not until the summer of 1945 that agreement 
was reached and the Eastman team's approach to 
what became known as "Amber" was chosen. 
That did not allow enough time to create an oper- 
ational machine. Amber was not to appear until 
the war had ended. 87 

(T0//0I) Eastman's proposed Amber was to 
be a photoelectric comparator, following in the 
traditions of Icky and Hypo. It had four scanning 
photocells that tested two 70mm repeatedly off- 
set and superimposed films against a coincidence 
threshold. It used the familiar analog circuits but 
with a sophisticated twist: both positive and neg- 
ative photosensing was used. 88 

(T8//8I) Amber was to have a. relatively large 
viewing field so that a long string of code could be 
tested instantly. It was designed to test eight hun- 
dred characters a second. The long field would 
give Amber speed and more. It increased the 
probability that "false hits" wouM be eliminated. 
More than that made Amber powerful; it had a 
very special way of implementing the weighting 
methods used to automatically identify probable 
"plaintext." The "dots" on one of its films repre- 
sented the weights through varying densities. 

£fiJ//3"F) Those variable densities called for 
much to be added to the older "dot" cameras 
Eastman had made for Icky and Hype. A special 
card reader and camera combination was devel- 
oped. Amber's camera was a sophisticated exten- 
sion of the light-bank system that had been devel- 
oped earlier in the war for machines like Icky. As 
the weight cards were read, one of twenty differ- 
ent voltages was applied to the tiny lamps to 
achieve the variable densities. Its developers 
knew that it would be a major chore to keep the 
system in tune, but that was inescapable. More 
than the camera was demanding. The data were 
so densely packed that extra can had tc be taken 
at every step of film preparat'er and develop- 

(Tfl//3fj In addition, the older analog sum- 
ming and threshold circuits had to be revised. But 
Amber was not to be a completely new machine. 
Only the changes to the older Icky and Hypo con- 
cepts that were absolutely necessary were incor- 
porated. Thus, Amber had much of the crudeness 
of the Icky. When Amber's films were placed in its 
Icky-like projector, if enough of the code groups 
had clear "dots," Amber would just stop. 89 

(TS//SJ} Amber's design was very demand- 
ing. It needed controlled humidity and correctly 
monitored ambient lighting. And its film trans- 
ports had to be much more precisely adjusted 
than Icky's or Hypo's. One reason for the narrow 
tolerances was that the very expensive master 
films containing the "weights" would wear out if 
there was the slightest friction. 90 Those master 
films were precious because they carried the crit- 
ical "weights" that were so labor intensive and dif- 
ficult to calculate. 

( T3//8I ) Amber's creators convinced "G" 
there was no alternative but to accept the great 
burdens the "master film" design required. The 
cryptanalysts knew that the preparation of each 
log-odd weight film called for millions upon mil- 
lions of multiplications. Those assigned to the. job 
feared they would never be able to keep up with 
the task. Although they had the tabs, including a 
special Multiplier, the job of preparing the long 
card decks to feed the special film-generating 
camera seemed overwhelming. A short twenty- 
three-character message needed almost two mil- 
lion multiplications for its weight film. 91 It took 
over 800 hours oftabulator and IBM electro- 
mechanical multiplier time to create atypical 
deck. And each of the different types of JN37 
messages needed more than a dozen of its own 
"probabilities" films. 92 Unfortunately, a hoped- 
for emergency project to create an electronic mul- 
tiplying machine could not be initiated, and "G" 
had to accept the prospect of thousands of hours 
of calculations to prepare for Amber's arrival. 93 
Perhaps it was for the best that Amber was deliv- 
ered just after the war with Japan was concluded. 

Page 188 


TO P 3C0R[:T//00 M INT//rcCL TO UOA, AUG, CAM, G B R A.v JJ NZU/X1 

CT3//GJ) Amber did not achieve all the goals 
set for it in 1945, but it was eighty times faster 
than the proposed digital electronic JN37 
machine. And it could be modified to perform 
other than the "weight" test. It did a simple 
round-robin test of every message in a group 
against every other 1,000 times faster than the 
NCR-Gray Comparators. But its 800 compari- 
son-a-second rate (the original goal had been 
35,000 a second) 94 meant that a typical run for 
the "37" attack might have, at the very least, taken 
twenty-four hours. 95 

( TS// S I) Because of Japan's defeat, there 
would be only two Ambers, 96 not the twenty 
Howard Engstrom had sought earlier in the year. 
Some remodeling was needed in 1947, 97 but the 
Ambers proved useful into the 1950s. 98 

(U) Dr. Bush, Your Best Friend Is Really the 

(TG//GI) The contract for Amber was not a 
signal that OP-20-G had regained its faith in pho- 
toelectric machines. Its frustrations with it and 
the Gray-NCR Comparator led to a belief that a 
long development cycle would be necessary 
before Bush's ideas could be turned into the pow- 
erful and reliable machines Joseph Wenger and 
Admiral Hooper had longed for in the mid-i930s. 
Although "G" added an electronic "rare-event cir- 
cuit" to the Gray Comparators and had Icky refur- 
bished, it did not return to Eastman or Gray for 
more machines during the war. 99 

( 6//SI) By 1944 the lack of orders from the 
navy led Eastman to consider reassigning most of 
its RAM team, leaving only the Amber group in 
operation. 100 Then the army saved the day. 

(S#8B Some of the SIS's engineers had 
become devoted fans of Bush's visions, and by 
late 1944 they were laying plans for a whole series 
of microfilm-electronic RAMs, a series that went 
beyond what the navy had once imagined. They 
even requested a statistical Rapid Selector: a 

machine that married microfilm with electronic 
counting. And by early 1945 their belief in film led 
to a request that a new camera be constructed to 
allow the army's Gray-NCR Comparator to 
become a film rather than a punched-tape 
machine. The SIS had experienced so many prob- 
lems with the punches that it was searching for 
any way possible to prolong the operational life of 
the Comparator. ,Q1 

(S//&E ) Although it never acquired all that its 
engineers desired, somewhat ironically it was the 
army, not the navy, that fulfilled Bush's dream of 
a "statistical" Rapid Selector. By 1945 the SIS- 
Eastman teams put electronic counting together 
with microfilm. And they continued on after the 
war to be the sponsors of the most far-reaching 
attempts to create film-based machines for crypt- 

( 8//0I)- Friedman's team had begun its 
romance with RAMs in early 1943 when the navy 
allowed the SIS to piggyback orders for a few 
machines onto the navy's contracts. loa Close to 
$200,000 changed hands very quickly for the 
purchase of IC plate devices, atetragraph tester 
and, later, a Gray-NCR Comparator. The first 
purchases were just that, purchases. The SIS 
played no role in the design of the machines. But 
after the RAMs began to arrive in late 1943, the 
SIS wanted a more powerful voice in machine 

(G//0I ) Within less than a year, a subcommit- 
tee was formed by those in "F" who had become 
strong advocates of the film-based devices. Many 
ideas for new RAMs emerged. Eastman soon 
began creating an SIS version of Hypo (their film 
Dudbuster) and an upgrade on the Tessie. There 
were more ambitious plans. The SIS was develop- 
ing ideas for a film version of a Slide Run 
machine, an Icky for the Fish traffic, and a special 
type of Amber. A budget request of $75,000 for 
research was approved. In addition, funds were 
allocated for the initial development stages of at 
least three new machines. 


Page 189 


QBfffft) Only one of those proposed machines 
was constructed, but some unexpected film 
RAMs appeared at the SIS during late 1944 and 
early 1945. "* 3 

(TS//Sf) One of those surprises was the result 
of lashing together the army's version of the 
35mm Icky and the electronic ,v±~>leis, used on 
the Gray-NCR Comparator. Only one ofthese 
unnamed machines was built, but it pleased the 
SIS analysts and encouraged them to make more 
modifications to existing devices. The army's 
Tessie, for example, was significantly upgraded. It 
was made more efficient, but, more importantly, 
it was made to automatically return to the point 
on the films where a sought-after complex code or 
cipher pattern was located. Furthermore, a new 
camera allowed it to use the "bhckcni"" test for 
brute-force searches. Later, a rr.:*re reliable card- 
to-film converter was requested. 104 

(U) The Great 5202 

(TG//0I) While the older RAMs were being 
updated, Eastman was busy with the SIS's major 
RAM contribution, the "5202." TV 5° 02 was the 
machine for the Fish system that had been rec- 
ommended in 1944. It became the mos" sophisti- 
cated and powerful of all the film RAM machines 
of World War II and after. In fa?t, the 5202 effort 
can be considered to be the major catalyst in 
keeping Eastman-Kodak together after the war. 
Although it was completed and sent to England 
too late to make its mark against Germany, the 
5202 was used throughout the 1940s ,<>5 and was 
used to attack the German Tunny ^ciphering 

CCS//SI^ Among the 5202's 1 danced features 
was its much-improved camera system. It took 
the light-bank principle far beyond the previous 
versions. Very important, it could pack the pat- 
terns generated by analogs of encryption 
machines much more densely than earlier mod- 
els. As important, the circuits al'ieri with the pat- 
tern gene-ators allowed great flexibility in select- 

ing the data transmitted to the camera. Creating 
complementary code patterns, for example, was 
very easy. The camera system went far towards 
solving a major problem of all the older film sys- 
tems: the great amount of time it took to gener- 
ate the films. 106 

ffSf/S-Q The 5202's heart, its reader, was also 
a technical improvement over the previous 
Eastman RAMs. Optics were improved In combi- 
nation with the dense packing on the films, the 
improved sensing systems allowed ten times the 
number of characters to be tested at once as on 
the other film RAM. The 5202's span ofsoo 
columns was impressive. That made 5202 a more 
robust cryptanalytic aid than the other RAMs 
with their relatively short viewing gates. 

(TS//SI) T he 5202 was also versatile. It could 

be used to "locate" desired patterns as well as to 
make Comparator-like counts on its electronic 
banks. It could hold as many as four films at a 
time; two of those were motor driven. 
Furthermore, its drive mechanisms were 
extremely fast and could step films in many dif- 
ferent patterns. 

< -¥G/f9f ) The unique feature of the 5202 that 
gave it the potential to be as valuable as the 
British Robinsons or Colossi was its ability to test 
two fields of data at the same time. That allowed 
it to perform the special cryptanalytic test it 
embodied. The 5202 could demand that no con- 
tradictions in two fields be found at the same time 
that ore or more "confirmations" were located. 
To dc that, the 5202 contained sensing and test- 
ing circuits that sought electrical balance among 
three photocell circuits. 107 

£ES/y$9 Although designed for the German 
teletypewriter problem, many different applica- 
tions were found for the 5202. It was used as a 
statistical dudbuster, for example. 

Page 190 



(U) Beyond the Comparators 

fT0//GI j Well before the 5202 was delivered, 
the SIS was drawing up plans for yet another new 
generation of 70mm film RAM. Their immediate 
target was to be Japanese code systems, but the 
machines were intended to be pathbreakers to a 
new era in microfilm devices. 

(TS//S ' I ) The first request was for a much 
enhanced Eastman version of Steinhardt's 
Copperhead I. That would allow high-speed 
searches through very long portions of text (hun- 
dreds of characters rather than only the thirty in 
Tessie). Next came a request for a new type of 
5202, one to perform isomorphic tests. Most 
important, according to the "F" group, was a film 
Slide Run machine with very advanced electron- 
ics. It was to be 100 times more powerful than the 
relay-based IBM versions. 

(T8//SI) The requests explained that each 
machine was urgently needed for attacks on 
Japanese weather and army codes. But the "F" 
group, to be credible, had to acknowledge that it 
might take some time to develop the new RAMs - 
perhaps too long given the signs that the Pacific 
war was winding down. To avoid losing their 
machines, they provided a thorough analysis of 
the role of film computation in SIS's future, hop- 
ing that even if the request for specific machines 
was rejected, research funding would continue. 108 

(S//SI) Emphasizing that it was crucial to 
keep the RAM group at Eastman together, "F" 
asked for enough money to sustain, at the least, a 
research effort in Rochester for several years. 
They admitted that the machines they needed 
were more of a challenge than the 5202, but they 
said the future ofSIGINT demanded new 
RAMs.' 09 

(U) The Machine That Wasn't 

(C//0I) There was one challenge that "F" 
group did not attempt to meet, a possible RAM 

that was not even mentioned during the war - a 
machine for traffic analysis. Neither the army nor 
the navy tried to create a machine for "data pro- 
cessing." The absence of massive fast memories 
and rapid input -output equipment meant that lit- 
tle attention was paid to creating a revolutionary 
data processing engine. A "data" machine had to 
wait until agency priorities changed and until 
computer readers, printers, and memories with 
capabilities far beyond those of the early 1940s 

(U) Notes 

1. (.TV/S-tj A glimpse into the complexity of the 
weighting methods is found in(¥S)-NSA CCH XII Z, 
Lt. A. H. Clifford to Lt. J. H. Howard, "Full Selector: 
operation 01: a four digit code group differences," 22 
February 1945. When a similar method was used 
against the Japanese weather systems, the required 
calculations to arrive at the weights for the "statistical" 
attack proved so numerous that the cryptanalysts 
asked that an electronic multiplying machine be con- 
structed, ("mffm) NSA CCH Series XII Z, R. A. 
Rowley, "Preparation of Weighting Film, Secondary- 
Stage Problem," OP-20-G, 2 August 1945. 

2. (T0//DI) The use of log weights to estimate 
whether or not plain language was appearing as the 
result of a decryption process was well established by 
mid-war. The Gee-Whizzer had been built around the 
idea. (S//SI) NSA CCH Series XII Z, Oral History 
Interview OH 04-82 with Samuel S. Snyder, 24 
February 1989. The very innovative Bulldozer (dis- 
cussed below) was aBombe version of the method. 
£CS} NSA CCH Series XII Z, CNO, CITS Paper TS-30, 
"Bulldozer Supplementary Manual," Navy Dept., 
Washington, November 1945. 

3. (TS//SI) NSA CCH Series VII Z, L. R. 
Steinhardt, "Possible Engineering Solutions for Full 
Selector Problem," OP-20-G-4-A5, 23 November 
1944. The "M" engineering group had kept in touch 
with all Allies' computer and electronic development 
projects of World War II. The failure to mention elec- 
trostatic memory' in the JN25 problem reports was 
perhaps not a result of ignorance but of a knowledge of 
both the primitive stage of development of the tech- 


Page 191 


nology and the unlikelihood that potentially fast elec- 
trostatic memories would hold large amounts of infor- 
mation. Useful for insights into memory technology of 
the era is James W. Cortada, Historical Dictionary of 
Data Processing Technology (New York: Greenwood 
Press, 1987). 

4T t 3 //64 » NSA CCH XTT Z, OP-20-G-4-A, 
"Electronic Matrices," 19 September ^44. 

5.fS) Such a "memory* was not uncommon at the 

6. (TSft**} NSA CCH Series XII Z Samuel S, 
Snyder, "Famous First Facts, NSA: ?a:'t 1, Pre- 
compiler Machine Cryptanalysis." 

7. (S//SI) NSA CCH XII Z, OP-20-G-4-A, 
"Electronic Matrices," 19 September 1944. 

8. (U) The highly significant Duenna project is dis- 
cussed below. 

9. (U) On the tube reliability question and the suc- 
cess of their use in the machine that : s generally held 
to be the "first'' electronic computer (although it had a 
special-purpose architecture), see Njncy Stern, From 
ENLAC toUNIVAC: An Appraise! of the Eckert- 
Mauchly Computers (Bedford, Mass: Digital Press, 
1981). The ENIAC had some 18,000 tubes and a very- 
high downtime because of that. 

10. (Si. Ideas for use of photocells and glass plates 
were suggested in some of the NDRC fire-control pro- 
posals. Those seemed impractical to Steinhardt. 

11. («//«J NSA CCH XII Z,OP-2C-G-4-A, 
"Electronic Matrices," 19 September 194-4, 

12. (U) Louis A Gebhard, Evohitir-n nfthe Naval 
Radio-Electronics and Contritmthns of the Naval 
Research Laboratory , (Washington, D.C.: Naval 
Research Laboratory, 1979), 326-8, also displays a 
veiy advanced German acetate tape system that was 
capture;! during the war. 

13. CSym NSA CCH Series XII Z, L. R. 
Stein hard t, "Additive Machines: F : "<criral Summary 
of," 27 November 1944. TSO Lt. A. H. CliTord to Lt. H. 
H. Howard, "Full Selector: operation on 4-digit code 
group differences," 22 February 19*3- ¥■)• NSA CCH 
Series XII Z, "Conferences at Da}ton," n -'.pril 1945. 

14. (U) OP-20-G and the engineers at SIS were not 
the only Americans searching for appio;;riate "com- 
puting technologies" to meet escalating demands. See 
for example, Chapters two and three, in Kent C. 

Redmond and Thomas M. Smith, Project Whirlwind: 
The History of a Pioneer Computer (Bedford, 
Massachusetts: Digital Press, 1980). 

15. tS77*^ NSA CCH XII Z, Conferences at 
Dayton," 11 April 1945. Interviews with Phil 
Bochicchio, June 1994. (U) W.W. Stifler, Jr. (ed.) 
High Speed Computing Devices: By the Staff of 
Engineering Research Associates, Inc. (New York: 
McGraw-Hill Book Company Inc., 1950), 346. 

16. (6#SD NSA CCH Series XII Z, "Conferences at 
Dayton," 11 April 1945. Lt. Reid was also important to 
the project. The same type of selector problem drove 
OP-20-G back to film machines. But the Eastman- 
Kodak "Amber," described below, was created for the 
Japanese weather systems. 

17. #} NCML had received a German magnetic 
disk sometime late in the war, but magnetic disks were 
already known to Americans. Like the computer itself, 
magnetic disk technology was "in the air." 

18. 0S#«) NSA CCH Series XII Z CNO CITP TS- 
32, "Mercury," Washington, D.C.: Navy Department, 
December, 1945. (R) NSA CCH Series XII Z, 
"Mercury," May 1953. 

19. (TS77*}) NSA CCH Series XII Z CNO CITP TS- 
32, '•Mercury," Navy Department, Washington, D.C., 
December, 1945. "fRi. NSA CCH Series XII Z, 
"MercUiy," May 1953. 

20. 3S//8+) NSA CCH Series XII Z, CNO, "Brief 
Descriptions of RAM Equipment," Washington, D.C.: 
Navy Department, October 1947, 14. 

21 CESj^Sl) Mercury also had an electronic circuit 
that cc-'jld set and record the "slides" the collator per- 

Z2.(rBfim NSA CCH Series XII Z, CNO, "Brief 
Descriptions of RAM Equipment," Navy Department, 
Washington, D.C.: Navy Department, October 1947, 

23. OPSjV*;) NSA CCH Series XII Z, "Office of 
Computers, List of Computers," nd. Mercury was 
retired ir March 1949. 

24. £¥&#«-) NSA CCH Series XII Z, "Report of 3 
January 1944 by Lt. L. Steinhardt: JN157 Rattler." 

25. gmffSf) NSA CCH Series XII Z "Report of 3 
January 1944 by Lt. L. Steinhardt: JN157 Rattler." 

Page 192 



26. - ffly/Oir NSA CCH Series XII Z, 
"Memorandum of 3 January 1944 From Lt. L. 
Steinhardt, Another Idea for Rattler," 5 January 1944. 

27. (T3//3I) OP-20-G-43, "FINAL REPORT, 
Project M-242, Rattler." 

28. (S#SQ NSA CCH Series XII Z, Inventories of 
RAM Equipment, 1945. (T0//DI) NSA CCH Series XII 
Z, "Office of Computers, List of Computers," nd. (U) 
NSA RAM File, CNO, U.S. Naval Communications, 
CITP TS-6 "Rattler," Washington, D.C., circa 1945- 
NSA NCML-CSAW Message File, October 22, 1943, 
Steinhardt to G, "Viper Design"; November 24, 1943, 
Ely to Desch, "Design of plugboard to automate Viper 
stecker analysis" and August 14, 1943, "Python to be 
shipped to Washington." 

29. {S//SQ NSA CCH Series XII Z, L. R. 
Steinhardt, "General Purpose Machine (SERPENT)," 
OP-20-GE, 29 September 1944. 

30. (U) It is important to note that Steinhardt did 
not mention the Robinsons, which were multitape 
machines. Perhaps that was because he had not been 
told of them. 

31. CS#«) NSA CCH Series XII Z, L. R. 
Steinhardt, "General Purpose Machine (SERPENT)," 
OP-20-GE 29 September 1944. 

32. (S//S*) NSA CCH Series XII Z, L. R. 
Steinhardt, "General Purpose Machine (SERPENT)," 
OP-20-GE, 29 September 1944, 4. "Rattan problems 
will demand a truly versatile machine in the early ana- 
lytic stages. One such problem (involving something 
like the proposed "Imagination Machine") is now cur- 
rent; this could be handled very nicely on Serpent." By 
the end of the war, a traditional type of analog machine 
was built for one of the Russian devices. The 
Americans called their machine "Ricky." 

33- (TS//SI) NSA CCH Series XII Z, OP-20-G 
"War Diary Reports: March 1, 1943-May 31, 1948," 
August 1945. 

34. (U) David J. Crawford, The Autoscritcher and 
the Superscritcher, forthcoming, The Annals of the 
History of Computing , NARA RG457, SRH-361. 
"History of the Signal Security Agency, Volume Two, 
The General Cryptanalytic Problems," 269-270. NSA 
RAM File, CNO, U.S. Naval Communications, CITP- 
TS-39, "Duenna Operations Manual," March 1946, 
and TS-20 "Bulldozer Operating Manual." 

35- fFS7/SI) NSA CCH Local Archive, "Army-Navy 
Descriptive Dictionary ofCryptologic Terms," Army 
Security Agency, February 1947, 131, defines scratch- 
ing as the testing of assumptions by examining its 
implications for jntradictions, eliminating those with 
contradictions, tht .1 "scoring" the remainder. 

36. CFS/r&* NSA CCH Series XII Z, "Tentative 
Brief Description of General Analytic Equipment for 
Enigma Problems, ' 26 March, 1945. f[4304] 9. 

37. (3S#&!-) NSA CCH Series XII Z, "Tentative 
Brief Description of General Analytic Equipment for 
Enigma Problems," 26 March, 1945. f[4607] f[4i42] 
129 f[4i49]- 

38. ePfr//Sl) NSA CCH Series XII Z, (S12008) 
Navy Dept., Office of Chief of Naval Operations, DNC 
(OP-20-G), RIP 425, "The American Attack on the 
German Naval Ciphers," October 1944 [sic], 129. (3=8)- 
NSA AHA ACC 17480, CNO CITS TS-17, "The N-800 
Bombe," Washington, circa 1946. (¥8) (S2568) NSA 
CCH Series XII Z, "Tentative Brief Description of 
General Analytic Equipment for Enigma Problems," 
26 March 1945. 

39- 0S#«r) NSA AHA ACC 35173, (CNO, CITS 
TS-49, "A Posteriori Remarks on the Cryptanalytic 
Aspects of the Bulldozer," Navy Dept., Washington, 
September 1946. (TS//SI) NSA CCH Series XII Z, OP- 
20 to King, Duenna. C E S//8 ft NSA CCH Series XII Z, 
Wenger to OP-20, to March 1945, "Statistical Bombe- 
Installation of. ' 

40. ( TS//S1 ) NSA CCH Series XII Z, OP-20-G, 
"Memorandum, Statistical Bombe-Successful 
Installation of," 10 March 1945. A statistical grenade 
was the first goal. (3S) NSA AHA ACC 35173, CNO, 
CITS TS-49, 'A Posteriori Remark on the 
Cryptanalytic Aspects of the Bulldozer," Navy Dept., 
Washington, September 1946. 

41. ( T S //S3 - NSA CCH Series XII Z, OP-20-G, 
"Memorandum, Statistical Bombe-Successful instal- 
lation of," 10 March 1945. 

42. (T0//SI) NSA AHA ACC 35173, CNO, CITS, 
TS-49, "A Posteriori Remarks on the Cryptanalytic 
Aspects ofthe Bulldozer," Washington: Navy Dept., 
September 1946, 3. 

43- (TS77ST) NSA AHA ACC 35173, CNO, CITS TS- 
49, "A Posteriori Remarks on the Cryptanalytic 
Aspects ofthe Bulldozer," Washington: Navy Dept, 


Page 193 


September 1946, pages 3 and 8 give the "weights" 
assigned to each letter based upon analysis of Enigma 

44- ( T5//3r ) NSA AHA ACC 35173, CNO, CITS TS- 
49, "A Posteriori Remarks on the Cryptanalytic 
Aspects of the Bulldozer," Navy Dept., Washington, 
September 1946, points out that the practical limit in 
the machine was a forty-five-letter . ipher because of 
the probabilities of Enigma wheel turnovers. 

45- (T 3//3r ) NSA CCH Series XII Z, CNO, CITS 
Paper TS-30, "Bulldozer Supplemental Manual," 
Washington: Navy Dept., November 1945. Because 
cipher-only attacks could produce so many false hits, 
using Bulldozer on tests for wheel order and stecker 
led to extraordinarily time-consuming print checking. 
The unknown stecker was the truly difficult problem. 
(TS) NSA AHA ACC 35173, CNO, CITS TS-49, "A 
Posteriori Remarks on the Cryptanah "k Aspects of the 
Bulldozer," Navy Dept., Washington, September 1946. 

46. (¥S//S*)" NSA CCH Series XII Z, file folder, 
"Monogram and RAM Panel Reports. 1945-1949." The 
Americans had not yet told the British about the 
Bulldozer in 1947. On the fears about the security of 
the ECM, (8)- NSA CCH Series XII Z,OP-20-G to 
Admiral King, "This may develop." 

47- ETS//SI) NSA CCH Series XTI Z, LeRoy H. 
Wheatley, "Cryptanalytic Machines in NSA," 30 May 
1953, indicates that the pluggable refertor was prima- 
rily a Yellow system machine. (TS//*l) NSA (CCH 
Series XII Z, IT. H.Campaigne and J 7. Pendergrass, 
"Second Report on Cryptanalytic Use o f High Speed 
Digital Computing Machines," OP-m-L, J) December 
1046, Appendix 1. shows that both the dermaii Army 
and Air Force were beginning to use the nev reflector. 
The OP-20-G cryptanalysts built Duenna on the 
assumption that each problem would take 5 x 10 to the 
sixth power tests. fS) NSA CCH Series XII Z, OP-20- 
GMF, "Report: Proposed Design for Drenna. Mark 
One," 25 February 1944. 

48. P&ffSTfNSA CCH Series XII Z, "Uncle 
Walter," circa 1945- 

49. {Sy&l} NSA CCH Series XTT 7., FAM list and 
Conference at Dayton, 11 April 1945 

$o.J2&Um NSA CCH Series XI' Z, LeRoy H. 
Wheatley, "Cryptanalytic Machine? in NSA," 30 May 
1953, "Duenna." (S} NSA CCH Series XII Z, OP-20- 

GMF, Report: Proposed Design for Duenna Mark 
One," 25 February 1944, 11. 

51. ( T6//SI ) The most complete descriptions of 
Duenna and its allied cryptanalytic process are found 
in (¥8) NSA CCH Series XII Z, CNO CITS Paper TS-39 
"Duenna Operations Manual," Washington, D.C.: 
Navy Dept., March 1946; C#fNSA CCH Series XII Z, 
CNO, CITS Paper TS-39 "Duenna, Theory Manual," 
Washington, D.C.: Navy Dept., July 1946; f&) NSA 
CCH Series XII Z,0P-20-GMF, "Report: Proposed 
Design for Duenna Mark One," 25 February 1944; NSA 
AHA ACC 25057, CNO CITS TS-39 "Duenna Electrical 
Circuits," July 1946. 

52. (T-StTW) NSA CCH "P" Collection Box CCO 67, 
RIP 608, CITS Paper TS-10/E-6, Enigma Series Vol. 6, 
Duenna," CNC-OP-20, January* 1946, contains the 
technical description of "G's" version of the scratching 

53- (TS//SI ) NSA CCH Series IV B-1-2, History of 
the Signal Security Agency, Volume Two: The General 
Cryptanalytic Problem, 268, implies that three differ- 
ent methods were developed by OP-20-G and SIS for 
the reflector problem and that the SIS attack and 
machines were significantly different from "G's." It 
also states that GC&CS's Giant machine embodied the 
SIS attack. 

54- G8T NSA CCH Series XII Z, OP-20-GMF, 
"Report: Proposed Design for Duenna Mark One," 25 
February 1944, 12. 

55. (JS#ei) NSA CCH "P" Collection Box CCO 67, 
RIP 608, CITS Paper TS-10/E-6, "Enigma Series Vol. 
6, Duem-ia," CNC-OP-20, January 1946, 6-3. Note that 
quite a different run time needed for Duenna to solve 
a problem is given in (XS//S1) NSA CCH Series IV B- 
1-2, "History ofthe Signal Security Agency, Volume 
Two: The General Cryptanalytic Problem," 269. It 
states that it took Duenna two weeks to test "all con- 
stantatiors" in a problem. The vast difference in time 
estimates is due to different definitions of "problem" 
and attack. 

E 6. PPS#S.I) NSA CCH Series XII Z, H.H. 
Campaigns and J.T. Pende'-grass, "Second Report on 
High Spe=d Digital Computing Machines," OP-20-L, 
18 December 1946, Appendix I. (S#») NSA CCH 
Series XTI Z, OP-20-GMF, "Report: Proposed Design 
for Duenna Mark One," 25 February 1944, contains a 

Page 194 



useful description of the Duenna menuing and of its 
printing system. The search and test logic is explained 
in, NSA AHA ACC 25057, CNO CITS TS-39 "Duenna 
Electrical Circuits," July 1946, 7-11. 

57. fJS#Sf) NSA CCH Series XII Z,(Si20o8) 
Navy Dept., Office of Chief of Naval Operations, DNC 
(OP-20-G, RIP 425, "The American Attack on the 
German Naval Ciphers," October 1944 [sic] £¥5f NSA 
CCH Series XII Z, CNO CITS Paper TS-39 "'Duenna 
Operations Manual," Washington, D.C.: Navy Dept., 
March 1946, shows how the commutators' sensing sys- 
tems were modified to fit the Duenna problem. 

58. (3//0I) NSA CCH XII Z, OP-20-G-4-A, 
"Electronic Matrices," 1-9 September 1944. f[4253] 

59. &Sjf$T) NSA AHA ACC 25057, CNO CITS TS- 
39 "Duenna Electrical Circuits," July 1946, 13, 32, 44. 

60. (U) David J. Crawford and Philip E. Fox (ed.), 
The Autoscritcher and the Superscritcher: Aids to 
Cryptanalysis of the German Enigma Cipher Machine, 
1944-1946," IEEE, Annals ofthe History of 
Computing , vol. 14, No. 3, 1992, 9-22. The 
Autoscritcher seems to have come into operation in 
early 1945, some months after Duenna had been 
brought to life. 

61. S S//M ) NSA CCH Series IV B-1-2, "History of 
the Signal Security Agency, Volume Two: The General 
Cryptanalytic Problem," 269. 

62. CIS^#3 NSA CCH IX.B.1.9, SSA "History of 
the Signal Security Agency, Volume Nine, History of 
the Development Branch," 10 February 1953, 76. 

63. W/ST) NSA CCH Series IV B-1-2, History of 
the Signal Security Agency, Volume Two: The General 
Cryptanalytic Problem, 269, gives the date of first 
operation as Christmas 1944. 

64. (T0//0Q NSA CCH Series XII Z, LeRoy H. 
Wheatley, "Cryptanalytic Machines in NSA," 
September 1954. 

65. ( TS//SB NSA CCH Series XII Z, LeRoy H. 
Wheatley, "Cryptanalytic Machines in NSA," 
September 1954. £ES} (S2568) NSA CCH Series XII Z, 
"Tentative, Brief Description Of General Analytic 
Equipment for Enigma Problems," 26 March 1945. 

66. (TS//SI ) One bank was required for each 
cipher-plain pair being tested. (3S)- NSA AHA ACC 
11254, "OP-20-G, "Army Autoscritcher," 29 March 

67. (U) Dc,\id J. Crawford and Philip E. Fox (<jg.), 
"The Autosci.tciiei and the Superscritcher: Ads to 
Cryptanalysis cf iht German Enigma, Cipher Machine, 
1944-1946,'' SEE, Annals ofthe History of 
Computing , ,ol. 14, No. 3, 1992, 12. 

68.-f¥977#) MSA CCH IX.B.1.9, SSA "History of 
the Signal Security Agency, Volume Nine, History of 
the Development Branch," 10 February' 1953, 78. 
Although the machine proved difficult to maintain, it 
was used against "E" traffic and, then, attached to the 
003 to function as a crib dragger for Swiss Enigma 
problems. fFS)- (£.2568) NSA CCH Series XII Z, 
"Tentative Brief Description of General Analytic 
Equipment f iu Enigma Problems," 26 March 1945, has 
an explanation o'ihe use ofthe "cups." 

69. (U) David J. Crawford and Philip E. Fox (ed.), 
"The Autoscritcher and the Superscritcher: Aids to 
Cryptanalysis of the German Enigma Cipher Machine, 
1944-1946," IEEE, Annals ofthe History of 
Computing , vol. 14, No. 3, 1992, 15. (S#8f) NSA AHA 
16899N, Army Service Forces, C. R. Deeter, "General 
Specifications and Technical Description: Super- 
scritcher," 13 January 1945. 

70. GBJ^/6iO NSA CCH IX.B.1.9, SSA "History of 
the Signal Secur ty Agency, Volume Nine, History of 
the Development Branch," 10 February 1953, 80. 

71. WW) NSA AHA 16899N, "Super-Scritcher: 
System and Circuit Details," points to the innovative 
ways the engineer's avoided many ofthe pitfalls of 
purely digital circuits. 

72. (Sftm NSA AHA ACC 16899N, Harry B. 
Smith, "Ring of Modified Eccles- Jordan Trigger 
Circuits," 12 January 1945. 

73- (U) Many ofthe electronic circuits were clever 
variants of digital designs, especially those designed to 
circumvent the need for hundreds of tubes to imitate 
or sense rotor signals. See £E8}- NSA AHA ACC 
16899N, Army Service Forces, David J.Crawford, 
"Frequency Sensmg in Rotor Outputs: Super-scrrtch- 
er," 17 January 1945- CCS//SH) NSA CCH IX.B.1.9, SSA 
"History ofthe Signal Security Agency, Volume Nine, 
History ofthe Development Branch," 10 February 
1953, 81. 

74- eW/WT NSA CCH Series XII Z, "Office of 
Computers, List of Computers," nd. 


Page 195 


75. (TS//SI) After the war OP-20- G did order an 
upgrade on 1CKY, and Eastman reworked Amber and 
H>po. £ES) NSA CCH Series XII Z,file folder, 
"Monogram and RAM Panel Reports, 1945-1949." But 
there was much discontent with film and the mainte- 
nance problems of the devices. 

76. (U) Samuel Eliot Morison, History of United 
States Naval Operations in World War U,v&ivne 
X1U, The Liberation of the Philippines, Luzon, 
Mindanao, the Visayas, 1944-1945 (Boston: Little 
Brown and Company, 1975), 59. 

77. CES#SO NSA CCH Series XII Z, OP-20-G 
"Analysis of Analytical Machine Attack on. IN-37," 24 
March 1945. 

78. (TS#»0 NSA, CCH Sena? XII Z, H.H. 
Campaigne, ".'N-37, Prospectus of Attack On," OP-20- 
G, 27 January 1944. 

79. CI2#») NSA CCH XTT Z ; "St.VJrfc;! Projects 
Needed: First Report of," OP-20-G4A, 30 April 1945. 

80. (TS//SI) NSA CCH Series XK Z,"0P-20-G: 
Analysis of Analytical Machine Attack on ./N-37," 24 
March 1945. (TO//0I) NSA CCH Series XII Z, OP-20- 
G4A, "JN-7 Strength of Additives : c ''oin a Two-Deep," 
5 September 1945- (Wt*3 NSA CCH Series XII Z, 
"Utilization of Available Climatologjed Data for JN-37 
PJ?Jn Text Estimates," OP-20-G-4A, 13 Jane 1945. 
&&ri$r) NSA CCH Series XII Z,"JN-37: Major 
Computation Needed for Machiic ^V'ght'".." OP-20- 
G4-A, 3 May 1945. 

81. (£^ NSA CCH Series XII Z, J. A. Skinner, 
"Proposal for Decoding Device," OP-20-G, 16 
February 1943. £[4022]. 

82. (S/jWJ NSA CCH Series XH Z, L. R. 
Sfcrnhardt "Additive Machines: Historical Summary 
of," 2-- November 1944. (W#Sf) OP-2U-C.-4-A5, 23 
Ncverrber 19.44, L. R. StekhvcL "Possible 
Engiutfdng Solut'ons for Full Selector Problems." 
OS^ST) NSA AHA ACC 26373, "-d ? Rovlett, 
"Report Eythe Subcommittee On "lie Application of 
Rapid AnaHtical Machinery t?the Solution of 
Enci'Dbered Code," 3 November 1944 (0//SI) 
Stehihardt, L.R., "Copperhead it(P:"GJect M-230) 
Final Report," 9 November 1944. tSVS* 1 NSA AHA 
15^5. John N. Seaman, "Memcnwdum for Major 
Edgerton, Li? ; ;,on with Navy #3, Us? <-.f Pr-jp on Jap 
Na>al Problems ofB II Type," 9 Juno :.Q44- 

83. (S/^S9 NSA CCH Series XII Z, J. H. Howard, 
"Summary of Conference on the '37* Machine With 
Eastman-Kodak and NCR Co.," OP-20-G, 9 June 1945- 

84. (-¥ 6/ /9t ) NSA CCH Series XII Z,"0P-20-G: 
Analysis of Analytical Machine Attack on JN-37," 24 
March 1945, part I Introduction, 8. 

85. CES^ST) NSA CCH Series XII Z,"0P-20-G: 
Anahsis of Ar.a'ytical Machine Attack on JN-37," 24 
March 1945. 

86. ( TS//SI) NSA CCH Series XII Z, "OP-20-G: 
Anarysis of Analytical Machine Attack on JN-37," 24 
March 1945, part I Introduction, 10. 

87. (&#m NSA CCH Series XII Z, J. H. Howard, 
"Summary of Conference on the '137" Machine With 
Eastman Kodak and NCR Co.," OP-20-G, 9 June 1945. 

88. fFSfT^f) NSA CCH Series XII Z, D. L. Noble, 
"Machine Weight Study for Proposed JN-37 Machine 
Part V OP-20-G-4-D, -'tjJiily 1945, and NSA CCH 
Series XII Z, D. I- Noble, "Machine Weight Study for 
Proposed JN-37 Machine Part II," OP-20-G, 2 August 
1945, give insights into how difficult it was to set and 
maintain the photoelectric system so that false hits 
would be avoided and tn.e hits tagged. 

8c. @8tfG& NSA CCH Series XII Z, LeRoy H. 
Whsatley, Cryptanalytic Machines in NSA," 30 May 

90. [T&tfSfJ NSA CCH Series XII Z, J. H. Howard, 
MN-£7 Macbi'i?: Report of Conference at Eastman 
Kodak Co. on 25 May 1945," OP-20-G, 26 May 1945. 

91. (JS#8fl NSA CCH Series XII Z, R. A. Rowley, 
"Preparation of Weighting Film, Secondary Stage 
Problem," OP-20-G, 2 August 1945. 

9:1, (TSyySQ-NSA CCH Series XII Z, R. A. Rowley, 
"Preparrtion cfWeighting Film, Secondary Stage 
Problem," 0P-2O-G, 2 August 1945. 

03. PPS7W NSA CCH XII Z, "Statistical Project 
Needed: ^irst Report of," OP-20-G-4A, 30 April 1945. 

9 J. Q&tf/rr) NSA CCH Series XII Z,"OP-20-G: 
AiieHsu of Analytical Machine Attack on JN-37," 24 
March to->5. 

95- C&fffi) NSA CCH Series XII Z, LeRoy H. 
Wheatley, "Cryptanalytic Machines in NSA," 30 May 
1953, NSA. CCH Series XII Z,List of Machines and 
Target?, circa 1945. Actual run-times were not given 
for "27" jobs on Amber so an estimate had to be made 
based upon s general idea of the speed of the machine 

Page 196 



and the degree of parallel processing built into it. The 
twenty-four-hour estimate was based on 800 compar- 
isons per second and 100,000,000 tests. This estimate 
of hours coincides with an official estimate fcr a 
round-robin of 1,000 500-letter messages for a gener- 
al coincidence test. 

96. (TS//3T) NSA CCH Series XII Z, "Office of 
Computers, list of Computers," nd. 

97. ffO//3Q NSA CCH Series XII Z.file folder, 
"Monogram and RAM Panel Reports, 1945-1949-* 

98. tT0//3 r) NSA CCH Series XI K, MN-37 
Machine Memoranda," and "Some Uses of Amber in 
Hagelin Attack," December 1949. 

99. &&^f&ff OP-20-G did give Eastman a signifi- 
cant contract for new models of ICKY and Hi'PO after 
the war. However, it took many years for the new 
models to be delivered. 

iOO.XS#8f) NSA AHA ACC 26373, SIS, "Miiutes 
of RAM Meeting," 19 February 1945. 

101. W/ST) NSA AHA ACC 26373, SIS, "Technical 
Paper, RAM," circa June 1945, 3, 5. The SIS consulted 
with the navy about the new RAM ideas. -(fr//9#-NSA 
CCH Series XII Z, OP-20-G, "SSA Proposal for 70mm 
Film I.C. Machine," 8 June 1945. 

102. (S#9f) NSA AHA ACC 26373, Frank B. 
Rowlett, "Report By the Subcommittee On the 
Application of Rapid Analytical Machinery to the 
Solution of Enciphered Code," 3 November 1944. 

103. CS^STJ Apparently, it was men such as Dale 
Marston who took the lead as Leo Rosen seemed to 
favor electronic versions of the special-purpose tab- 
relay machines that IBM was building for the Agency. 
fjS//3fl NSA CCH Series XII Z, Robert O.Femer, 
"Rapid Analytic Machinery Needed for Research," 
June 3, 1943. On the RAM plans, ( S /'/S Ij NSA AHA 
ACC 26372, SSA, "Rapid Analytical Machinery," circa 
October 1943. W/6t) NSA AHA ACC 26373 Frank B. 
Rowlett, "Two Copies of Report on Rapid Analytical 
Machinery," 3 November 1944. 

104. (58#M) NSA CCH Series XII Z, "The Status 
of RAM," circa June 1945- (S//S1) NSA AHA ACC 
26373, Chief, "F" Branch, "RAM Equipment," 29 
March 1945. 

105. (W//W ) NSA CCH Series XII Z, "The Status 
of RAM," circa June 1945, *6#S») NSA AHA ACC 
26373, SIS, "Technical Paper, RAM," circa June 1945- 

(fl//0I) NSA CCH ACC 26373, 'Twenty-Fourth-RAM 
Report," 1 May 1945. 

106.JCES//SQ NSA CCH Series XII Z,"The Status 
of RAM," circa June 1945, 7. 

107. (S#») NSA AHA ACC 29373, SIS Chief T 
Branch, "Request for RAM Equipment," 23 March 
1945, 7- 

108. (S//SI) NSA AHA ACC 29373, SIS Chief "F" 
Branch, "Request for RAM Equipment," 23 March 
1945. 0£#m NSA AHA ACC 26373, Frank B. 
Rowlett, "RAM in Future Cryptanalysis," 3 May 1945. 

109. (S//9f) NSA CCH Series XII Z, OP-20-G, 
"SSA Proposal for 70mm Film I.C. Machine," 8 June 
1945, commented on the army's request. 


Page 197 


75- ff S//SB After the war 0P-20-G did order an 
upgrade on 1CKY, and Eastman reworked Amber and 
Hypo. (XS) NSA CCH Series XII Z,file folder, 
"Monogram and RAM Panel Reports, 1945-1949." But 
there was much discontent with film and the mainte- 
nance problems of the devices, 

76. (U) Samuel Eliot Morison, History of United 
States Naval Operations in World War H, v^hine 
XIII, The Liberation ofthe Philippines, Luzon, 
Mindanao, the Visayas, 1944-1945 (Boston: Little 
Brown and Company, 1975), 59, 

77. (W#gQ NSA CCH Series XII Z,0P-20-G 
"Analysis of Analytical Machine Aita^k on JN-37," 24 
March 1945. 

78. (TS#S*) NSA, CCH Series XII Z, H.H. 
Campaigne, ".IN -37, Prospectus of Attack On," OP-20- 
G, 27 January 1944. 

79- CIS#S*) NSA CCH XI 1 2, "Sta:i<,fcl Projects 
Needed: First Report of," OF-20-G4A, 30 April 1945, 

80. (TS//SI) NSA CCH Series XII Z, fc 0P-20-G: 
Analysis of Analytical Machine Attack onJN-37," 24 
March 1945- (T0//D1) NSA CCH Seri?s XII Z, OP-20- 
G-1-A, M'N-7 Strength of Additives *h>m a Two-Deep," 
5 SppJember 1945. (SSt¥#) NSA CCH Serias XII Z, 
"Utilization of Available Climatologjoal Data for JN-37 
Phin Text Estimates," OP-20-G-4A, '-3-Jjne 1945- 
CtS^/ST) NSA CCH Series XII Z, "JN-37: Major 
Computation Needed for Maohiio V"i'ght i ' ; ° OP-20- 
G4-A, 3 May 1945. 

81. &U&) NSA CCH Series XII Z,J. A. Skinner, 
"Proposal for Decoding Device/' OP-20-G, 16 
February 1943. £[4022]. 

82. (SA<6$ NSA CCH Series XTI Z, L. R. 
Stemharot "Additive Machines: Historical Summary 
of," 27 November 1944- CSjW) *-:*;- Q-4-A5, 23 
November 1944, L. R. Sterdnrct "Possible 
Engiht'e.dng Solutions for Full Selector Problems." 
G67YST) NSA AHA ACC 263/3, ?■"<'•.'• ? Rowlett, 
"Report By the Subcommittee On 'lie Application of 
Rapid AnaHrical Machinery to the Solution of 
Enciphered Code," 3 November 1944 (0//SI) 
Stein hardt, LR., "Copperhead 11 (Project M-230) 
Final Report," 9 November 1944. tS'/S*) NSA AHA 
15 ?5, John N. Seaman, "Memorandum for Major 
Edgerio/y Li? ; ;,on with Navy #3, I T s? of P.r-.rr on Jap 
Naval Problems o f B II Type," 9 June -.944. 

83. &W NSA CCH Series XII Z, J. H. Howard, 

"Summary- of Conference on the '37' Machine With 
Eastman-Kodak and NCR Co.," OP-20-G, 9 June 1945. 

84. mm NSA CCH Series XII Z,"OP-20-G: 
Analysis of Analytical Machine Attack on JN-37," 24 
March 1945, part I Introduction, 8. 

85. CES#3T) NSA CCH Series XII Z,"0P-20-G: 
Anah sis of Ana'ytical Machine Attack on JN-37;' 24 
March 1945. 

86. ( TS//SI) NSA CCH Series XII Z, "OP-20-G: 
Analysis of Analytical Machine Attack on JN-37," 24 
March 1945, part I Introduction, 10, 

87. ($tm NSA CCH Series XII Z, J. H.Howard, 
"Summary of Conference on the '137 Machine With 
Eastman Kodak and NCR Co.," OP-20-G, 9 June 1945. 

88. «FS#S-P5 NSA CCH Series XII Z, D. L. Noble, 
"Machine Weight Study fox Proposed JN-37 Machine 
Part V OP-20-G-4-D. "9 July 1945, and NSA CCH 
Series XTI Z, D. L Noble, "Machine Weight Study for 
Proposed JN-37 Machine Part II," OP-20-G, 2 August 
1945-, giro insights into how difficult it was to set and 
maintain the photoelectric system so that false hits 
would b? avoided and trie hits tagged. 

go. eWffl* NSA CCH Series XII Z, LeRoy H. 
Wheatley, Ciyptanalytic Machines in NSA," 30 May 


90. (*KSjW NSA CCH Series XII Z, J. H. Howard, 
MN-£7 Machine: Report of Conference at Eastman 
Kodak Co. on 25 May 1945," OP-20G, 26 May 1945. 

91. CXS#Stf NSA CCH Series XII Z, R. A. Rowley, 
"Preparation of Weighting Film, Secondary Stage 
Problem," OP-20-G, 2 August 1945- 

9:1 (3$#S*)-NSA CCH Series XII Z, R. A. Rowley, 
"Preparri'on cf Weighting Film, Secondary- Stage 
Problem," D -2o-G, 2 August 1945. 

93. ffiS//«f; NSA CCH XII Z, "Statistical Project 
Needed: First Report of," OP-20-G-4A, 30 April 1945. 

9t. (FvfffT) XSA CCF Series XII Z,"0P-20-G: 
Ato.Hsl, of Analytical Machine Attack on JN-37," 24 
March :o.-;5. 

95. CEBtfSt) NSA CCH Series XII Z, LeRoy H. 
Wherttey, "Ciyptanalytic Machines in NSA," 30 May 
1953, N"?A CCH Series XII Z,List of Machines and 
Targ-ls, :irca 1945. Actual run-times were not given 
for "37" jobs on Amber so an estimate had to be made 
based upon ageaeral idea ofthe speed ofthe machine 

Page 196 



and the degree of parallel processing built into it. The 
twenty-four-hour estimate was based on 800 compar- 
isons per second and 100,000,000 tests. This estimate 
of hours coincides with an official estimate fcr a 
round-robin of 1,000 500-letter messages for a gener- 
al coincidence test. 

96. (T3//3I) NSA CCH Series XII Z, "Office of 
Computers, List of Computers," nd. 

97. ff3//39 NSA CCH Series XII Z,file folder, 
"Monogram and RAM Panel Reports, 1945-1949." 

98. ( - ¥ //Sr) NSA CCH Series XI K, MN-37 
Machine Memoranda," and "Some Uses of Amber in 
Hagelin Attack," December 1949. 

99. ffS^§r)- OP-20-G did give Eastman a signifi- 
cant contract for new models of 1CKY and HYPO after 
the war. However, it took many years for the new 
models to be delivered. 

ioo.XS#Sf) NSA AHA ACC 26373, SIS, "Mnutes 
of RAM Meeting," 19 February 1945. 

101. <S#ST) NSA AHA ACC 26373, SIS, "Technical 
Paper, RAM," circa June 1945, 3, 5. The SIS consulted 
with the navy about the new RAM ideas. 4&-/i$$-NSA 
CCH Series XII Z, OP-20-G, "SSA Proposal for 70mm 
Film I.C. Machine," 8 June 1945. 

102. CSyye*) NSA AHA ACC 26373, Frank B. 
Rowlett, "Report By the Subcommittee On the 
Application of Rapid Analytical Machinery to the 
Solution of Enciphered Code," 3 November 1944. 

103. CS>¥WJ Apparently, it was men such as Dale 
Marston who took the lead as Leo Rosen seemed to 
favor electronic versions of the special-purpose tab- 
relay machines that IBM was building for the Agency. 
rS#»3 NSA CCH Series XII Z, Robert 0. Ferner, 
"Rapid Analytic Machinery Needed for Research," 
June 3,1943- On the RAM plans, (S//SI) NSA AHA 
ACC 26372, SSA, "Rapid Analytical Machinery," circa 
October 1943. W/S** NSA AHA ACC 26373 Frank B. 
Rowlett, Two Copies of Report on Rapid Analytical 
Machinery," 3 November 1944. 

104. CB8#9I) NSA CCH Series XII Z,The Status 
of RAM," circa June 1945. £S//SJ» - NSA AHA ACC 
26373, Chief, "F" Branch, "RAM Equipment," 29 
March 1945. 

105. C*9#6J) NSA CCH Series XII Z, "The Status 
of RAM," circa June 1945, £S#S1) NSA AHA ACC 
26373, SIS, "Technical Paper, RAM," circa June 1945. 

(3//0I) NSA CCH ACC 26373, Twenty-Fourth-RAM 
Report," i May 1945. 

loe.iXSr'rSf) NSA CCH Series XII Z,The Status 
of RAM," circa June 1945, 7. 

107. (S#») NSA AHA ACC 29373, SIS Chief "F" 
Branch, "Request for RAM Equipment," 23 March 
1945, 7- 

108. (S//SI) NSA AHA ACC 29373, SIS Chief "F" 
Branch, "Request for RAM Equipment," 23 March 
1945. (X S // S Q NSA AHA ACC 26373, Frank B. 
Rowlett, "RAM in Future Cryptanalysis," 3 May 1945, 

109. @ffm NSA CCH Series XII Z, OP-20-G, 
"SSA Proposal for 70mm Film I.C. Machine," 8 June 
1945, commented on the army's request. 


Page 197 

This page intentionally left blank 

Page 198 


Chapter 7 

(U) The Magic Continues 

(U) Would History Repeal Itself* 

(U) Hindsight might lead one to think it 
would have been easy to predict the futuie of the 
United States and its signals intelligence services 
after World War II. The Cold War and res conse- 
quences for America's military, industrial, and 
academic life seem "natural." The rise of Big 
Science, the development of a new type of univer- 
sity, and the growth of a massive intelligence 
establishment intertwined with public and pri- 
vate high technology institutions appear histori- 
cally inescapable. Even the close relationship 
among Western nations, as reflected by the for- 
mation of NATO, and the exceptional cooperation 
between the intelligence agencies of Britain and 
the United States appear to have been foreor- 

(U) But in 1945 all of that was in the future, 
and much of it came as a surprise to the nation 
and to the leaders of the American communica- 
tions intelligence community. For contempo- 
raries, the future was uncertain. No one imagined 
that America was going to build a mul tibillion- 
dollar intelligence bureaucracy that seemed to 
have a life of its own. In fact, for those in the 
cryptanalytic organizations in early 1945, there 
were signs they might return to the isolated and 
have-not world of the 1930s, an era when 
American politicians condemned "reading other 
gentlemen's mail." The concerns about STGTNTs 
future had some foundation. Communications 
from the White House were less than subtle 
reminders that even the Soviets' communications 
should be treated as sacred. 

( TS//SI) None of the codebreakers wanted a 
reversion to the attitudes and inadequate budgets 
of the 1930s, but a few thought that some degree 

of isolationism might be useful. They reasoned 
that c. WcV tcixt'p monies flowing to SIGINT 
would be to ena World War II's dependency on 
Great Britain. A handful of influential men in the 
intelligence community suggested that the ties to 
Britain's ci^ptanalytic organizations be cut. With 
America's policymakers having to depend only on 
"G" and the SIS, they reasoned, there would be a 
decreased chance of their allocations being 
slashed. 1 

fTS/-/-^) Those beliefs were significant and 
contributed to the rejection of one of the most 
generous offers Britain ever made to the United 
States. The British had come to consider their 
attack on tl:e Fish systems as their great cryptan- 
alytic achievement. They saw it as their intellectu- 
al and, because of Colossus, their technological 
triumph. They also viewed Fish as an example of 
the cryptanalytic systems of the future. They fore- 
saw the day when most systems would rely upon 
baudot-teletype, not Morse transmissions. 

0FS//S9- When they offered the United States 
one of their Colossus machines in the summer of 
1945, and more than hinted it would be followed 
by the cryptosecrets it required, they were 
shocked toleg.rn their offer was rejected. They 
found it difficult to understand why and so did 
some Americans. Joseph Wenger wanted a 
Colossus in the United States, but the head of the 
technical arm of the SIS, Frank Rowlett, thought 
it unwise to accept a machine that was so complex 
and so expensive to maintain. 2 Combined with 
the desire for autonomy, Rowlett's objections 
were convincing. 

(U) The rejection did not mean that American 
SIGINT planners were sure of their technological 
future. Men like Joseph Wenger had good reason 


Page 199 


to fear the consequences of peace. World War II 
had not yielded permanent solutions to most of 
the problems that had frustrated him and Bush 
during the 1930s. The goal of creating a perma- 
nent intelligence-gathering capability had not 
been achieved and, certainly, the grand institu- 
tional dreams of Admiral Hooper had not been 

(U) And as the war was ending, the achieve- 
ments that had been made in the previous four 
years were threatened. It appeared that OP-20-G 
and the SIS would have to struggle to improve if 
not just maintain their powers. And they would 
have to do it alone. America seemed to be return- 
ing to a prewar institutional profile. The corpora- 
tions, the universities, and the military were 
pulling apart. 

(U) Big Science seemed about to be torn 
down. Vannevar Bush's NDRC had been a gener- 
ous but only temporary big brother for the mili- 
tary and the aspiring research universities. In 
1945. when NDRC funds were being cut off, no 
one was sure that academics wanted to or could 
continue to supply intellectual and technical solu- 
tions to the military. 

(U) A greater cause for worry was the indica- 
tions that the unity of purpose among the large 
corporations and the government was about to 
end. The relationship that had developed during 
the war seemed to be too costly to maintain. 
Industry's desire to return to high-profit work 
appeared to be impossible to overcome. No mili- 
tary agency could guarantee the high and consis- 
tent rates of return needed to persuade major cor- 
porations to continue to devote themselves to 
responding to military needs. 

CSS) In addition, the private sector gave few 
indications that it would support expensive long- 
term investigations and developmental work that 
might compensate for the termination of the 
wartime research programs. Many promising 
technologies that had appeared during the boun- 

ty years of the war stood in danger of being 
ignored as military budgets declined. It seemed 
unlikely, for example, that universities would 
continue the type of research into radio wave 
propagation that had proven so useful to OP-20- 
G's interception program. There also seemed lit- 
tle hope that corporate programs would lead to 
the advanced demultiplexing equipment that the 
army and navy so desperately needed to tap into 
the modern transmission systems of their tar- 
gets. 3 

(U) The prediction of the future of the federal 
scientific establishment was not comforting. The 
best forecast seemed one of a return to the lean 
1930s. A weak National Bureau of Standards, a 
small navy Office of Research and Inventions in 
competition with a feeble ONR, and an army 
communications research program confined to a 
hungry Signal Corps might be all that the com- 
munications/intelligence agencies could look for- 
ward to. 

(U) During the last months ofthe war, the 
gloomy predictions seemed to be coming true. 
Although the post-World War II situation would 
be infinitely better than during the 1920s, the 
army and navy cryptanalytic agencies would not 
have the partners, the resources, nor the autono- 
my they had during the preceding four years. It 
was not easy for them to continue to be techno- 
logically innovative, and, although they had a few 
more years of cryptanalytic "magic," they soon 
found it nearly impossible to meet the challenges 
of their most important cryptologic adversaries. 

(U) What There Wasn't 

(U) Like the story of OP-20-G and the SIS 
during the 1930s, the history of computers, auto- 
mated cryptanalysis, and data processing in the 
SIGINT agencies between 1945 and the mid- 
1950s can be understood only in the context of 
what was not available to America's codebreak- 
ers. What wasn't there extends to much more 
than hardware. The institutional structure need- 

Page 200 



edfor adynamic response to technologically 
sophisticated mathematical/cryptanalytic con- 
frontations did not exist. As important, World 
War II had not led to any great cryptanalytic 
methods revolutions. Although there had been 
much effort, mathematical cryptanalysis was 
something that still had to be created. 

(U) There is a long list of other particulars. 
Chief among the "missing" were the modern com- 
puter and an industry willing to provide the spe- 
cial types of devices required for codebreaking 
and traffic analysis. 4 In 1945 the modern elec- 
tronic digital computer was still a wish whose 
outlines were just being drawn. There were no 
commercial firms that were investing large 
amounts in its development, and academia, 
though willing, showed little sign of being able to 
carry the financial burden of bringing the univer- 
sal computer to life. 

(U) And many of those who saw the new elec- 
tronic computer slowly emerging from university 
and corporate centers had very serious questions 
if that "serial" machine could ever have the power 
needed to conquer cryptanalytic enemies. The 
critics of the emerging general-purpose computer 
desired machines with more complex architec- 
ture, ones that relied upon multiple processors 
and parallel action. Such machines seemed 
unlikely to appear on their own, however. There 
were no indications that any outside group would 
even attempt to bring such alternative architec- 
ture to life without direct and massive support 
from the codebreaking community. 5 Worse, few 
companies seemed willing to take the money OP- 
20-G and the SIS did have to spend on automa- 
tion. Both agencies experienced great difficulty 
before the 1950s in finding responsible contrac- 
tors who would commit to building the latest gen- 
eration of special-purpose electronic devices. 

(U) Intellectual resources were also absent. 
The in-house mathematical groups that the two 
agencies had created had to be vastly reduced in 
size at the close of the war. There was no ready- 

made substitute. The American universities had 
not yet reestablished ways to link themselves to 
secret military projects, and the armed services 
and their major contractors had yet to invent the 
"think tank." The Rand Corporation, with its abil- 
ity to allow academics to change into strategic 
planners, remained only a thought in the mind of 
the most aggressive air force generals. 

(U) Institutional power was declining. 
Neither OP-20-G nor the SIS was sure that it 
could maintain the degree of autonomy granted 
to it during the war. The "G" section and the "F" 
group stood under the threat of losing their free- 
dom to design machines and to select who was to 
build them. 

(U) Despite an immediate postwar generosity 
that extended the life of some programs, there 
were indications in early 1945 that signals intelli- 
gence might have to remain passive, only waiting 
on the sidelines while, hopefully, someone else 
made the great technological leaps needed to 
match advances in code and cipher making. 

(V) Signs of Some Appreciation 

(U) However, during 1945 there was some 
encouraging news. The SIS's and OP-20-G's 
wartime achievements had made them a few very 
good and very influential friends in the military. 
From the most important generals and admirals 
came words of praise for Ultra. With a little con- 
vincing by advocates such as Joseph Wenger, that 
praise was turned into promises by America's 
leaders to provide at least some of the resources 
needed to maintain and, perhaps, improve com- 
munications intelligence capabilities. 

(U) Although budgets were slashed and work 
forces seriously reduced, the SIS and OP-20-G 
were granted more than should have been expect- 
ed given their treatment in the 1930s. The fifty- 
percent reduction in the amount of IBM equip- 
ment in 1946, for example, still left the agencies 


Page 201 

top MCRETOenMiirnwFi toiisa aii^ pam rrp A M n M 7i/ f Y i 

with 300 machines. That was tens of times more 
than what was on hand at the start of the war. 6 

(TS//SI) The SIS was allowed a staff of 1,500, 
and "G" retained some 700 people - infinitely 
greater numbers than had been allocated to the 
services during the 1920s and 1930s. That gave 
some hope that although the United States had 
not yet decided to take responsibility for policing 
the entire world, a signals shutdown was unlike- 

(U) Significant for the history of computers 
and cryptanalysis were the numbers of "techni- 
cal" slots allocated to each agency. Both had 
approximately five percent of their staff approved 
to work on advanced technological and scientific 
matters. Those ratios were maintained during the 
postwar period, with the navy gaining a slight but 
important edge over the army in the numbers of 
high tech employees. 8 The technical sections 
were not large enough to support huge in-house 
production capabilities, but they were capable of 
pursviing some research and creating smaller 

(U) More than numbers of people indicated 
the value military leaders placed on centralized 
SIGINT. Although no firm promises could be 
made in 1945, hints were dropped that none of 
the cryptotechnologies would be abandoned. 
Whether the favored technology was the IBM tab- 
ulator-relay combinations or the electromechani- 
cal components in Madame X, they would be 
given a chance to develop into more powerful ver- 
sions of what had emerged during the war. 

(U) The most tantalizing hint concerned the 
possibility of an aggressive new Rapid Machine 
program. During late 1944 and 1945 the army and 
navy engineers were allowed to draw up some rel- 
atively long-term plans for very advanced 
machines. Different groups had their favored 
approaches. Many in the agencies wanted to con- 
centrate on extending the reach of the workhorse 
tabulating equipment. Others sought continuity 

through extending the reach of the Bush 
Comparator. A few continued to have faith in the 
type of analog and microfilm devices that had 
been built at Eastman-Kodak. Of course, there 
were vocal advocates for turning digital electron- 
ics into operational machines. 

fES-) There was a consensus on the role of 
SIGINT in developing all the technologies: "G" 
and the SIS should play an active part in bringing 
to life any new hardware. That was almost 
inescapable. There was much to be done and few 
on the outside willing to do it. There was no gen- 
eral-purpose electronic device that was a great 
advance over Bush's old Comparator, and even 
the mundane, such as input and output technolo- 
gy, remained at the early levels. 9 

(U) More MA GTC: Cryptanalysis Continues 
as Before 

( T3//SI) The chance to finally build analytic 
machines that would put America ahead of the 
technology of encryption seemed favorable dur- 
ing the first postwar years. One reason for that 
was America's continued cryptanalytic successes. 
The machines it had in hand were providing intel- 
ligence. In several instances "G" and "F" had to 
hurry to finish new machines, but, in general, the 
agencies did not have to concentrate, as during 
the war, on emergencies. For three years after the 
end of World War II, there was every indication 
that the kind of triumphs that had been achieved 
during the early 1940s would continue. As impor- 
tant, even the most dangerous ofthe nation's 
adversaries had decided to refrain from creating 
military crises. Success gave the agencies time to 
look ahead. 

(¥ g//0ij The impressive work ofthe 
American SIGINT agencies continued on after 
the Japanese surrender; remarkably, it even 
seemed to be improving. Every type of code and 
cipher used by every important nation was or 
seemed about to be conquered. Despite the intro- 
duction of more sophisticated methods and 

Page 202 



machines, some 70 percent of the systems of the 
minor nations were "readable" in 1946 and 1947, 
and over half of those of the three new major tar- 
gets were, to a significant degree, open to the 
United States. 10 

(TG//SI) With help from the British, some of 
the most critical and well defended of the Soviet 
codes and ciphers were yielding information. 
There was every reason to believe that what had 
been accomplished against the Germans and 
Japanese would be repeated against Russia and 
its allies. Progress seemed inescapable. By 1947, 
for example, engineers at the army and navy cen- 
ters could build relay analogs of some of the 
important Russian cipher devices, just as they 
had built analogs after Japan's Purple machine 
had been penetrated. 11 

CE S// S T ) The systems being entered were 
important. It appeared that, as in World War II, 
America would be able to gain critical intelligence 
through the interception and analysis of a rela- 
tively small number of messages, ones which 
would reveal the strategic thinking of the political 
and military leaders of Russia and its client 
states. 12 During the first postwar years, old 
attacks and machines were doing quite well. More 
than operator errors and busts were leading the 
Americans and British into the Soviet additive 
codes, [ 


I Although Russia's new 
Hmachine proved stubborn, 

there was hope that a persistent search using tra- 
ditional tools would uncover its wiring and its 
indicator system. 13 

(G//0I) That image of a consistent SIGINT 
future shaped the nature of the immediate post- 
war machine development programs at the 
army's and the navy's Washington headquarters. 
Both agencies sought cryptanalytic, not "data," 
machines. The goal in the first postwar years was 
to create new generations of machines to break 

complex codes and ciphers. Other possible 
sources of information, including traffic analysis, 
were left to older machines and methods. 

(TS//6I) Something besides that cryptovision 
shaped the initial postwar plans. It was thought 
that, unlike the years of the Atlantic U-boat crisis, 
no massive and emergency machine construction 
program would be needed. 14 

(U) A Cryptanalytic Future: Architecture 
and Ambiguity and Budgets 

(U) Everyone saw the future of "G" and the 
SIS as centered on the traditional cryptanalytic 
function. Traffic analysis, direction finding, and 
analysis of enemy clear text might play roles but 
very small ones. But all was neither secure nor 
settled, especially in 1945 when plans for both 
immediate and long-term machine development 
were being created. Although there was an agree- 
ment about general target priorities for the 
immediate postwar era, the exact nature of the 
cryptosystems and machines that might be con- 
fronted was not predictable. 15 

£¥&fflsl) Some trends were evident, however. 

J His company's products would proba- 

bly be used by every nation. The second trend was 
towards the adoption of on-line machines. Many 
nations were adopting the type ofbaudot- 
teleprinter and multiplexed systems that had 
formed the backbone ofthe German Fish net- 

(T3//3f) In addition, there were at least con- 
cerns that Enigma-like wired rotor machines 
might reappear, and it seemed likely that super- 
enciphered code systems would not vanish with 
the fall of Japan.The formidable one-time pad 
systems that seemed unbreakable when correctly 
used were known to be a favorite of many diplo- 
matic corps. 

EO 3.3(h)(2) 
PL. 86-36 


Page 203 

PiL. 86-36 


(TS//6I) But nothing about target systems, 
; with one exception, was so certain, nor the devel- 
jopmeht>pf the underlying technology for analytic 
Imachines so predictable, that a rush to create a 
;host of nevv special-purpose devices was justified 
An 1945. l6 There was no reason to build a series of 
; high-tech single-purpose, single-system mach- 
ines, except "'for, some of those manufactured by 



fe dealt with through existing relay, even tabula- 
tor equipment, as could many of the remaining 
additive systems, such as those used by Soviet 
]agencies. Not enough was 

known in 1945 about the new teletype-Baudot 
devices to warrant the construction of expensive 
special-purpose contrivances. As well, since the 
promises of the discovery of effective pure math- 
ematical methods had yet to be fulfilled, launch- 
ingUnto an expensive search for anew "calcula- 
tion" machine seemed unjustified. 

ffij// S I) Only one type of machine, anew 

I ; \ [ deserved special technolo gical attention 

atthei end of the war. It was the'fj 

series! The Americans knew it was going to be 
used; for high-level systems by many important 
nations; they had enough knowledge of its inner 
workings to challenge it; and they were confident 
they eould pick the correct attacks. 17 As a result, 
although many new machines would be built, the 
only adventurous special-purpose RAM planned 
for a s pecifi c problem at the end of the war was 
for thf~T 

(U) The Enigma Is Dead (We Think); Long 
Live the 

ffS//H) The first step towards anew 
machine program at"G" and the SIS was to sort 
out what WWII machines should be abandoned. 
All but a score of the navy Bombes were 
destroyed, and the million dollar Madame Xwas 
taken apart. The Duennas, the other elaborate 

Bombes, and the electronic Scritcher were kept, 
however. 19 It was thought they could be used 
against other rotor machines or the few Enigmas 
that might be brought back into use. All the 
devices that had been effective against the 
I I were retained. They were 

refurbished and readied to attack diplomatic, 
civil, and military traffic from around the world. 20 

(TS//SI) The Bombes and the other «E 

analogs were useless against 

Page 204 


-EO 3.3(h)(2) 
PL. 86-36 


C E8// 8 i j The investment needed to conquer' " 
the morel \ seemed 

reasonable. And success would provide invalu- 
able returns. For example, with Germany and^ 
Japan defeated, stripped oftheir colonies,^ 

ffSffST) The potential rewards from invest- 
ing in attacks a gainst oth er nations' Hagelins 
were even greater. j ^ras "the" manufactur- 

erof cryptodevices, and every nation, business, 
and bank in any part of the world had to rely upon 
his products. From South America to Northern 
Europe to Arabia, if automated encryption was 
b eing inte rcepted, it was probably the product of 
a£ \ T here was one other very important 

aspect of the 

one that helped 

launch the Amer ican cryptanalytic attacks against 
it. Thel 

jmachines could be purchased on 

the open market and their inner workings closely 


(U) A Hangover from Another Time 

(T0//8Ij One of the first postwar |_ 


RAMs had a strange beginning. It dated to when 
the navy still thought it would have to launch a 
bloody invasion against Japan. 22 "G's" mathe- 
maticians needed a large and fast digraph count- 
ing machine to attack various Japanese systems. 
The device had to be much more powerful than 
the creaky Mike. Although it would be expensive 
to build, "G" approved a request for the "counter." 
But the machine was almost canceled when 
Japan surrendered. Fortunately, the mathemati- 
cians were able to convince Engstrom and 
Wenger to continue the project. They agreed that 
despite the about-to-appear second Freak at 
Arlington Hall, the navy needed a universal 

counting device for its postwar missions. It would 
help, they realized, in the initial mathematical 
studies of unbroken systems and machines. 23 

6FS//0I) With promises of financing in hand, 
machine designers were consulted about techno- 
logical options. It was decided that the original 
proposal to build a machine to handle an alpha- 
bet greater than sixty-four characters was 
impractical. Also quickly rejected was the idea of 
basing the counting machine on a film-analog 
.combination as had been recommended earlier. 

(T8//gij The "G"NCML team explored other 
possible technologies and then made a surprising 
decision. Although "G" was filled with advocates 
of electronics, its engineers decided to let the 
army's Freak carry the risks of building an "elec- 
tronic" counter. Whether condensers or tubes 
were used, said the NCML crew, too many of 
them would be needed for a useful digraph 
machine. With a sixty-four by sixty-four matrix 
and the counters required to handle up to 999, 
the number of components, they concluded, 
would be too great. There would have to be more 
than 3*997 counters, each needing three posi- 
tions. Temperamental electric or electronic parts 
meant too many errors and too much "down 
time." For the navy's engineers, old-fashioned 
reliable electromechanical counters were the only 
viable alternative. 

6ffi//9Jr)" They were close to gaining the final 
approval of the cryptanalysts; then someone cal- 
culated the speed of the machine if it used the off- 
the-shelf industrial counters. The device would be 
incredibly slow, as sluggish as the old frustrating 
Mike. There was a standoff. The codebreakers 
wanted a fast machine, but the engineers would 
not accept the responsibility of an electronic 
device. After much wrangling, they arrived at a 
compromise. The engineers decided they would 
take the responsibility of designing custom-made 
mechanical counters that were fast enough to 
please the cryptanalysts. 


Page 205 


EO 3.3(h)(2) 
P.L. 86-36 

G ES//0I J After outlining the new counters, the 
engineers sought a contractor. Unfortunately, 
"G's" two largest World War II contractors, NCR 
and IBM, did not want to take the project. "G" 
had nowhere to go, and the machine was put on 
hold during much of 1945 and 1946. 

(TS// 8*) "G" waited until its "captive corpora- 
tion," Engineering Research Associates, was 
formed before it put any more effort into the 
counter-project. But once ERA agreed to take the 
contract, "G was sure it would soon have a useful 
and reliable machine.That was a rather naive 
assumption, however. 

(T8//SI) What became known as Alcatraz did 
not appear until 1950, had about half the power 
originally planned, and was much more expen- 
sive than expected. Once in operation, it threw 
technological tantrums. It had problems with its 
large printer, and the expense of maintaining the 
machine led to Alcatraz's very early retirement in 
1954- 24 

P Alcatraz 
(U) Mrs. O'Malley's Wayward Son 

(S// 8 I) Another of the special devices OP-20- 
G thought it had to have in order to deal with its 
postwar targets became, arguably, the largest 

electronic imitation of an adding machine ever 
built. Filling half a room with vacuum tubes, 
relays, a special card reader constructed by the 
IBM spin-off Commercial Controls, and a tabula- 
tor's printer, O'Malley was one of postwar "G's" 
earliest and most challenging projects. 

(S//SI) Something like O'Malley had been 
desired since 1942, but it was the growing backlog 
of messages I """I and 

some technological advances that sparked its 
final design in 1947. O'Malley was the cryptana- 
lysts' special and very grand version of what IBM 
had introduced at the end of 1946, an electronic 
multiplying machine. O'Malley had to be special 
and had to go beyond IBM's offering because it 
was to accomplish what Bush and Wenger had 
agreed was not achievable in the mid-i930s, the 
automation of the advanced version of an ICtest, 

(0//0I) The cryptanalysts' "Chi" was a very 
close relative of the Chi Square test. Chi Square is 
a now familiar statistical 
method for determining if 
two distributions came 
from the same "universe." 
It seems commonplace and 
unsophisticated today, but 
in the 1940s, especially 
because of the tedious cal- 
culations it needed, Chi 
seemed very advanced. It 
was one of the most sophis- 
ticated ways to identify 
cipher alphabets produced 
by the same key. 25 

(S//SI) As it was used in 
cryptanalysis, "Chi" was 
computationally demand- 
ing. The frequencies of each letter in one cipher 
text had to be multiplied against the correspon- 
ding frequencies in another text: then the prod- 
ucts had to be summed and used in the algorithm 
which determined whether or not the sum was 

Page 206 

TOP TE"PETr r> "' tlT '" :>cl ™"™ «i ^ V 1 1 1 1 nnn tmHSUm 

feO 3.3(h)(2) 
PL. 86-36 

top ocorccTV/oo Mi hm/rccL to uoa, auo, oa n , ODR, AND NZU/X 1 

likely to have been produced by chance. To isolate 
probable "same key" messages called for testing 
each text against all the others. That meant thou- 
sands of multiplications, hundreds of additions, 
and dozens of evaluations for the simplest 

(3//8I) The manpower-starved "G" realized 
that if the World War II victories against ! 

devices were to be continued, it had to have an 
ultra-powerful machine for accumulating the fre- 
quencies, calculating the sums of cross-products, 
identifying the "significant" sums, and displaying 
the results. Given the exploding Wo rkload of 
those charged with keeping up with th ej 
generated traffic, it was decided that an electron- 
ic high-speed multiplier was worthy of an enor- 
mous financial investment. 

(G//SI) The new navy contractor, 
Engineering Research Associates (ERA), 
was asked to build a super-fast machine 
that could recognize individual letters, 
tally them, and then perform all the thou- 
sands of multiplication and summations 
needed to identify those cipher messages 
whose letter frequencies "correlated." 
And it was asked to produce it quickly. "* 

(S//SI) That was a demanding 
request in the technological context of 
1947. ERA was expected to produce a 
machine more powerful than what had 
been developed by the world's largest 
and most advanced computing machine 
company: the IBM 603 electronic calcu- 
lator, which had been introduced in late 
1946. 27 

(S//SI) Within a year and one-half, 
the men in St. Paul were able to build an 
electronic "calculator" several orders 
more complex than the IBM device; but 
the need to shift attention to the produc- 
tion of another machine, one for critical 
SIGINT fire fighting, led toO'Malley 

being far below original expectations. To save 
design and production time, O'Malley became a 
"get the job done" machine. It was stripped of 
many of its intended powers so that it could be 
out into operation as soon as possible. 

($//3f) Many of the interesting technical 
challenges inO'Malley's original specifications 
were avoided. First, it was decided that, as in the 
1930s, recognizing and tallying letters were 
chores best done by older methods and technolo- 
gies. Separate machines would prepare the letter 
frequency counts. O'Malley was deprived of even 
more functions. It did not include the circuitry 
needed for an automatic test for the significance 
of the results. All the summed cross products 
were printed, leaving the analysts with the need 
to do much hand calculation. And O'Malley even 


top oconrn/oo MiN TtfncL to uoa, auo, cam, opr, amp h elotm 

Page 207 

10 3.3(h)(2) 
P^L 86-36 


heeded help reading its input. Given the available 
I/Q technology, a special machine had to be con- 
structed to punch tallies into IBM cards in a spe- 
cial format and the standard IBM readers had to 
be reconfigured. 

($> //0 fr Without that jury-rigged equipment 
and its special dual teletype tape readers, 
O'Maliej would have taken more years to con- 
struct. *G" could hot wait, for example, for mag- 
netic drums that could have provided a high- 
speed means of presenting two, streams of data. 
The quick fixes to the old card and tape technolo- 
gies had to be accepted. 

(0//0I) O'Malley became part of the tradition 
of the relay box and tabulator combinations of the 
NC Machines of World War II. Like them and the 
new IBM 603, it w$s a special-purpose (although 
electronic) calculation box inserted between tab- 
ulator-teletype input and output equipment. The 
inside of the machine also reflected the time pres- 
sures under which it was constructed. It certainly 
did not break any new logical ground; it was a 
direct imitation of a decimaKbased electro- 
mechanical calculating machine. It multiplied by 
following the traditional method of repeated 

(S//Sf) Such conservatism did not mean that 
O'Malley was a minor accomplishment, however. 
Commercial firms, such as IBM) Remington- 
Rand, and UNIVAC, based their early electronic 
offerings on the same philosophy of following 
known architecture and logic. Ring-counter, dec- 
imal machines were the norm. Binary devices 
were the challenging exceptions that were very 
slow to appear on the market or even in the labo- 
ratory. 28 

(«/#»)■ And for its time, O'Malley did its cal- 
culations quite rapidly. It was able to form and 
sum as many as thirty-five cross products in one 
and one-third seconds. It could have done its 
work much faster, but it was limited by its lack of 
memory and its very slow printer. 

(&//Q¥ ) O'Malley proved useful for more than 
its original limited operational job. When 
O'Malley was asked to serve the needs of research 
mathematicians, its tubes and relays were per- 
suaded to pretend they could, for example, invert 
matrices. 29 

(3//GB But all in all, economy and a need to 
obtain the final equipment as soon as possible 
meant that O'Malley was a compromise that had 
a short life. Perhaps it was its recurring I/O prob- 
lems that led to O'Malley's untimely and perhaps 
embarrassing end. It was allowed a relatively 
peaceful retirement twenty months after its birth. 

(S//J?J> The Grand 


(TS//St-> O'Malley and Alcatraz were not the 
heart of the planned postwar attack on the 

machines. A very special device had been 
uhdef consideration well before Germany's 
defeat. But the central 


Hecate, had to wait until "G" formed its captive 
corporation, ERA. 

( TS// 61) Hecate was more than two years in 
the making, arriving in Washington in 1948. Built 
with the help of the ex-OP-20-G engineers, such 
as John Howard, Hecate was a combination of 
the old and new. 

(3 $//0 f ) But even borrowing from the past 
did not make Hecate easy to complete. It was a 
huge and expensive combination whose cost esti- 
mate of $86,000 became a delivery price of 
almost $250,600. 3 ° Its price tag, the cost of five 
World War II Bombes, was justified by results, 
however. It and its sister p roduced a constant 
I I for more than a 

decade. 31 Hecate was, in fact, one of the electron- 
ic marvels of its time. 

(TG//SI) Hecate contained some true 
advances, such as its four hig h-speed elec tronic 
'ritvg s" that imitated the in the 

rThfiv ran at a very respectable 200 KC 


Page 208 



(U) Ifecate 

(100,000 trials a second). And Hecate's scritcher- 
type capability of eliminating branch searches 
speeded processing enormously. 32 

(T3//3I) To save construction time, electron- 
ics were employed only where demands for speed 
presented no alternative. Hecate might have been 
even more innovative, but it was called out of the 
workshop for immediate operational needs. Most 
of the machine was composed of the familiar and 
reliable plugboards and relays, and its essential 
logic was based upon familiar cryptanalytic-engi- 
neering approaches. 

(TS//SI) Hecate was not a digital computer 
and did not calculate. It was an analog "crib" 
machine with some "digital" components. like 
the Grenades, it used a short ten- or twelve-letter 
stretch of suspected plain text to identify "starting 
points." And it was not a complete processor. 

Much independent statistical analysis of a system 
was needed before Hecate could find the initial 
settings for a t ransmission. And its power was 
limited t o just 

( T9// 01) T o use Hecate the cribs were set with 
dials; then a message was read, letter by letter, 
into Hecate's relay memory through a standard 
tape reader. When the relay memory had the 
required number of letters in it, the electronic 
rings were run through their positions until the 
parallel flows of electricity through the plug- 
boards and relay circuits signaled that all the crib- 
plain pairs and ring positions were consistent, or 
that all the wheel positions had been tested. At a 
hit, the machine stopped and dials indicated the 
position of the "rings" and the place of the crib in 
the message. The dials were used because there 


EO 3.3(h)(2) 

Page 209 


P.L. 86-36 
EO 33(h)(2) 

was still no printer that was fast and reliable 
enough to compete with hand notation. 

- CTO//0IJ - Each of Hecate's offset letter tests 
took approximately three seconds, excluding the 
time an operator needed to write down the hit 
positions. It took a total of approximately twenty 
minutes to run a 500-letter message. 33 

(T8//SI) Hecate was reliable and effective, 
but, like the Bombes, she had some serious defi- 
ciencies. Because of the nature of her test, circuit 
completion, Hecate could produce too many pos- 
sible answers. Each had to be examined by a lim- 
ited workforce. To reduce the list of possibilities 
required difficult-to-find, very precise, and error- 
free c ribs. Worse, Hecate could be used again- 

(T S // S I) Hecate's limitations were recognized 
while it was under construction, but the machine 
was needed so badly that it could not be aban- 
doned nor radically altered. 

(TS//6I) OP-20-G had no solution for the 

of intelligence. The go-ahead was given for the 
development of another special-purpose device, 

<3 C//0I - ) Warlock I 

Page 210 



(U) Hecate's Impressive Competitor 

(TS//SI) Warlock was much more adventur- 
ous and costly than Hecate, so much so that it 
took some four years to design and manufactur- 
er 34 j t was physically big as well as expensive. It 
was so large that it had to be kept at the ERA fac- 
tory in Minnesota. Warlock turned St. Paul into 
something of a remote operations center. 35 

( rfi3//S f j Warlock cost more than $soo,ooo 36 
because it called on the sophisticated cryptanalyt- 
ictest that the Americans had first used in 
Bulldozer, automatic plain language recogni- 
tions. 37 Automatic recognition was demanding in 
its own way; it called for very, very long cribs. 
Fortunately, they did not have to be of a specific 
nature. To avoid the "false hits" that came from 
short and weak cribs, Warlock used some seven- 
ty-five or more ciphertext letters. All of them were 
needed just to eliminate much of the handtesting 
Hecate runs called for. 

(T6//GI) Warlock was a major engineering 
feat because it was very difficult to turn high- 
speed plain language recognition into electronics. 
Warlock was piled full ofthe latest electronic 
tubes and circuits. With the cipher text in place, 
super-f ast "digital" electronic wheels sped 


weighted each 

resulting letter according to its language frequen- 
cy, then summed all the results in parallel. An 
electronic threshold-testing component decided 
if the settings and a sample ofthe selected plain 
text should be printed. 

(TS//SI) One hundred thousand tests a sec- 
ond were performed, which meant that the plain 
text could be fully ex amined in twenty minutes. 38 
To run throug h the 

so quickly called for over 6,000 
tubes, crystal diode matrices, trays of relays, flex- 
ible electronic matrices, some binary circuitry, 
the familiar plugboards and relays, and even a 
magnetic drum. 

(TS//SI) The first Warlock w as a mach ine 
that could attack just one or a few sys- 

tems, but its 1953 version was able to do more 
ucLduac 01 iia aovdnceo electronics. riexiDie 

machines. For example, Wa rlock) was asked to; 
penetrate the mysterious 
machine that served all of" 

£TC//SI - ) - Although Warlock shared many cir- 
cuits, components, and ideas with the general- 

purpose computer, 
device with much 
Bulldozer. Although 

it was I a special-purpose 
ofthe I internal logic of 
yery effective, it was a/one- 
function machine, something many in the intelli- 
gence community, especially those in charge of 
budgets, disliked. 

(T3//9I) Both thd money managers /and the 
engineers wanted something else, a multipurpose 
cryptanalytic computer, one that could perform 
any or all the cryptoattacks.: Such a/machine 
would never sit idle waitifigj for messages from a 

particular system, nor 
dinosaur like Madame 

would it become a useless 
X.t° ; / 

(U) The Unwersal RAAfjs 

38//Sf) Many in QP-2p^G and the SIS want- 
ed to start designing such a universal RAM in 
1945. They were not abstract types; their RAM 
was to be tailored to operational cryptanalytic 
needs. Those advocates fori the universal RAMs 
refused to wait until aimathematical revolution 
transformed cryptanalysis or a technological rev- 
olution made the genet-all-purpose programmed 
computer a competitive! dijyptqtool. 

x (T&//0I) For those jlfn favor of universal 
RAMs, extending the general cryptanalytic tech- 
niques that had proved jjfo/valuable during the 
war seemed the only reasonable path for "G's" 
postwar machine prograiti/ 41 But they wanted to 
avoid the waste that wenfalong with creating spe- 


P.L. 86-36 
EO 3.3(h)(2) 

Page 211 

iup bhtKfc i ticmmimtiWEL 1 uoa, auo, qa h , oon, and hzl//x i 

cial-purpose machines. Given the unknowns 
about future cryptotechnology and the need to 
maximize research and development funds, the 
wisest choice for them was to create a machine 
that could perform all the major cryptanalytic 

ffG//0ft Those major functions fell into a few 
broad categories. The most important of the cryp- 
to-techniques were based upon either locating 
repeated patterns, tallying massive numbers of 
letter patterns, stripping possible cipher and rec- 
ognizing plain text, or performing some form of 
"exhaustive searching," such as done by the 
scritcher machines. 

( T0//0I ) The universal RAM was not to be a 
super-calculator for advanced mathematical cal- 
culation, or a direct analog of a cryptosystem, nor 
one that could be called a data processing 
machine. And although it was agreed it would be 
digital and electronic, it was to be something very 
special and unique to the cryptanalytic communi- 

(TS//8B In 1945 and early 1946, both agen- 
cies made a commitment to find, if possible, their 
own versions of one great multipurpose cryptan- 
alytic engine. 42 

(U) The Illusive Matrix 

(T0//0I) The call for the universal RAMs 
became tied with the search for an electronic 
"matrix." The universal machine needed it and so 
did a new type of Bombe, one that could attack 
any type of rotor enciphering machine through a 
Turing-like analog test. There was also a demand 
for an electronic matrix that would serve as the 
heart of all the more digitally oriented dedicated 
machines of the future, whether they were for 
wired-wheel or additive attacks. 

(T9//3I) The concepts of the matrices were 
not well formed in 1945. In some instances a 
matrix was described as being high-speed memo- 

ry, in others as an electronic version of a switch, 
and instill others as an analog of an encryption 
wheel. But whatever the purpose, existing tube 
technology made any matrix design very difficult 
to construct. The problem that had halted the cre- 
ation of an electronic Bombe in 1942 continued 
after the war. The matrices demanded too many 
tubes to be practical. 

( - RS// SI) Asa result, much of the "matrix" 
effort was concentrated on developing multipur- 
pose tubes and other basic components. That 
research became essential toother projects, 
including the search for the multipurpose crypt- 
analytic machine, 43 

Cftj //QI> In addition to the hunt for the elec- 
tronic "wheel" for the universal RAM, both serv- 
ices had special-purpose uses in mind for an elec- 
tronic matrix. Many in the navy wanted, as soon 
as possible, an electronic version of the valuable 
but none-too-well-behaved monster, Mercury, 
and the army desired a vastly improved version of 
its "look-up" devices, the Slide-Run machines. A 
few wanted an electronic super Bombe that could 
tackle many different machine systems through a 
Bombe-like test. 

£FS} But the demands for a universal machine 
continued. And soon its outline became clear. It 
would be something quite different from the pro- 
grammable general-purpose computer. 

(U) It's a Nice Idea, Dr. von Neumann, But... 

CCS} As "G" and the SIS focused on their ver- 
sions of one great device during 1945 and 1946, 
they came to quite similar concepts of a single 
machine that could perform all the general 
attacks that had proven so valuable during the 
war. Both had visions of a "computer" that 
performed IC tests, crib-dragging, locating, addi- 
tive stripping, and weighted plain language test- 
ing. The near-universal machines began to 
be assigned a generic name, a "reconfigurable 

Page 212 


lUHbfctiKblJJlUM I NO/REL T OU&A, AU3, CA N , QDR, AND N ZU/X 1 

(T0//0I ) Neither service was able to build its 
ideal "reconfigurable" machine during the 1940s 
because of institutional barriers, the primitive 
state of some of the underlying technologies, and 
emergencies that called for energies to be devoted 
to special-purpose machines. But "G" and the SIS 
went far towards defining a powerful and unique 
cryptanalytic computer architecture through 
their Goldberg and Sled projects. 

fTO//0I j There were differences between the 
content and progress of the two grand dreams, 
Goldberg at "G" and Sled at the SIS, 44 but at their 
beginnings they shared many fundamentals. 

£ Ki//flff Neither of the proposed all-purpose 
machines was conceived of in terms of the archi- 
tecture of the modern digital computer. Their 
designs were very different, for example, from 
what was emerging out of the ENIAC/EDVAC 
projects at the University of Pennsylvania. They 
were not single processor, serial, binary, and pro- 
gram-driven machines, the type that later became 
characterized by the term "von Neumann archi- 

(T0//DI ) Such an architecture seemed very 
inefficient to the cryptanalysts. Well into the 
1960s there were engineers and cryptanalysts 
who remained committed to the idea that the dig- 
ital, serial, single memory, program-driven "von 
Neumann design" for computers was an inappro- 
priate model for codebreaking. 

(TS//SI) Perhaps that was because their con- 
cepts of computers were problem, not abstrac- 
tion, driven. The proposed army and navy 
machines were not born out of considerations of 
how to solve any possible logical or mathematical 
problem. Neither was intended to be a universal 
logic or mathematical device. They were to be 
extensions of the hardware and methods that had 
evolved at the agencies during World War II. 

(TS//SI) The first source of inspiration for 
them came from the developments in the tabula- 

tor sections. Both agencies had invested in the 
creation of very efficient relay attachments for 
their tabulating equipment. The special IBM 
"boxes" became heroic in the eyes of the operat- 
ing cryptanalysts and their machine room allies. 
Each ofthose minicomputers, some ofwhich 
were much larger than the tabulators, performed 
a special function. The Slide-Run attachments, 
for example, stripped additives, then searched a 
dictionary of high-frequency code groups. The 
navy's NC tabs were also built from a wide range 
of functional relay boxes. 

(TS//SB- Another source of inspiration was 
the RAM program's faith in electronics and film 
and tape inputs. There had been many thoughts 
of enlarging the powers of the Comparators and 
the IC machines through the use of additional cir- 
cuits, ones that could be accessed through con- 
venient plugboard "programs." Putting the count- 
ing abilities and message-offsetting abilities of 
the Comparators together with the locating pow- 
ers of the Copperheads and the weighting capa- 
bilities of Amber seemed a possibility. 

GE S//S -I) Thus, it was a small evolutionary 
step to the central idea of the postwar Sled and 
Goldberg machines: embody each ofthe major 
cryptanalytic functions in separate hardware 
packages; create a central switching mechanism; 
tie the packages together in any desired configu- 
ration through the switch; and hook it all to free- 
standing input/ output mechanisms. 45 

(TS//fl i) With a stock ofthe specialized 
"boxes," the agencies could instantly create any 
desired cryptanalytic engine. The cryptanalysts 
would not have to wait for two or more years for 
a traditional type of special-purpose machine to 
be built; expensive machines would not become 
useless if an adversary changed his system; and 
the machine rooms would not be cluttered with 
devices that were used only a few hours a week. 

( TS//SI ) In 1945 there seemed, in fact, no 
alternative to such a machine. It appeared so nat- 


Page 213 


ural at the time that it did not have to be justified 
through a comparison with other possible archi- 
tecture. Because the "von Neumann" idea was rel- 
atively unknown in 1945, it was only later that the 
supporters of the multipurpose cryptanalytic 
machine concept justified their ideas through 
contrasting them to the universal serial comput- 
er. But when they did, they outlined an argument 
that has had a long life within the SIGINT com- 
munity. The arguments against the von Neumann 
design have continued for fifty years. 

(T0//0I> Bythe late 1940s, men in both "G" 
and the SIS were pointing out how their linked- 
box architecture would allow parallel processing, 
the incorporation, whenever desired, of analog 
computing, and thus much, much faster process- 
ing. They explained that the von Neumann design 
would always be too slow because it had only a 
single processor to do everything. To be useful, 
that processor had to be driven by an outside pro- 
gram, step by step by step. Hundreds, perhaps 
thousands of ticks of a clock had to go by before 
the most simple ofcrypto-functions could be 
completed. Nothing else could be done until the 
program cycle was finished. 

(T9//3I) However, the special boxes, 
arranged in the right manner, had the potential to 
be hundreds of times faster than the single 
processor device. Hardwired functions would 
always be performed in fractions of time it took to 
read and execute programmed instructions. And 
with the "boxes," while one function was being 
completed, another could be performed. As 
important, with a set of function boxes on hand, 
no one would have to wait the months, or, as it 
turned out, the years it might take to write a com- 
plex program for a von Neumann machine. 

( T0//0I) Most advocates of "reconfigurable" 
machines agreed on other things in 1945. There 
was a commitment to use and, if need be, create 
new components. The functional packages should 
be built with advanced electronics, if at all possi- ' 
ble, and with new input and output equipment. 

Only electronics could make the machines fast 
enough to perform the cryptanalytic tests; and for 
many of those tests, only new I/O devices would 
allow the electronics to work at optimum 
speeds. 46 

(U) Faith without Institutions: Slides, Sleds, 
and Skates 

OS// S3 In 1945 "G" and "F" did not realize 
how difficult it would be to follow through on 
their pledges to create a "universal" cryptanalytic 
machine. Who would be willing to build some- 
thing just for the SIGINT community? Who 
would pay the extra costs that necessarily came 
when standard, commercially produced equip- 
ment was rejected? Who could be trusted with the 
secrets that were embodied in the special devices? 

GB9//STD Although "G" and the SIS both spon- 
sored programs to create reconfigurable non-von 
Neumann machines at about the same time, and 
although the two agencies were required to coor- 
dinate their efforts, 47 their responses to those 
questions were quite different. The navy began its 
project immediately, but did not stay with all the 
original intentions for its Goldberg. The army 
eventually produced a machine that fit quite well 
with the original architectural vision, but it was 
almost a decade before its Sled emerged. 

(U) Among the many reasons for the different 
patterns one stands out: the navy allowed "G" to 
create, as will be described, a company that had 
the skills and the mandate to begin work immedi- 
ately. In contrast, the cryptanalysts and engineers 
at Arlington Hall were left dependent upon the 
vagaries of yearly budget allocations and the will- 
ingness of commercial corporations to subject 
themselves to what might become very unprof- 
itable projects. 

(T0//0I) But the army's project did not begin 
with a cloud over it. Leo Rosen's mid-1945 plea to 
extend the reach of the tabulator-relay combina- 
tions at Arlington Hall received a warm reception. 

Page 214 



The conceptual outlines for what was later called 
Sled were in the making by the end of the year, 
and there were signs that enough men would be 
left in his "F" section to design if not build the new 
all-purpose machine. Not all the functions that 
were to be turned into hardware were agreed 
upon, but there was little opposition to the idea 
that at very minimum, Sled would have to per- 
form the cribdragging and dictionary lookup 
functions of the invaluable Slide-Run machines of 
World War II. More ideas were contributed as to 
what functions should be included and, as impor- 
tant, how they could be designed to meet the goal 
of having, when desired, plugboard-programmed 
parallel processing. 48 

(8//SI) The evolving ideas even attracted the 
cryptanalysts and engineers at"G," especially 
those who had worked with the tabulator and NC 
machines. Already mandated by the government 
to coordinate as much research and development 
as possible, "G" and the SIS agreed they should 
work together on Sled. Each looked forward to 
having a Sled with electronic components within 
a short time. 49 

(T6//SI) However, the program ground to a 
near halt. Not enough resources and equipment 
remained in the postwar "F" section, and a con- 
tractor could not be found to turn rough ideas 
into specifications and hardware. Specifically, 
IBM was more than hesitant to accept the respon- 
sibility for the Sled program. 

( T0//0I) IBM had been a good friend to the 
SIS, and so many of its bright engineers had 
learned ofcodebreaking techniques during the 
war that IBM had seemed the logical choice to 
turn the "reconfigurable" idea into innovative 
machinery. There was also a factor of technologi- 
cal continuity. IBM had created the NC and relay- 
box devices that were to be paralleled in electron- 
ic form, and it was the manufacturer of the only 
efficient equipment for the cards that had become 
the standard "memory" in the tabulator rooms at 
both agencies. 

(T0//0I) IBM was such a clear choice that it 
seemed to have been the only one thought of by 
the SIS group advocating the Sled architecture. 
But IBM was not in the mood to take on such 
work after the war. Its management was not even 
extending a welcome to the requests from the 
Bureau of Ships for more NC machines including 
a critically needed one to record punch card data 
onto microfilm and then, if desired, reverse the 


( 3Ki//GI) IBM's rejection of the army and 
navy's overtures created intense shock at the SIS. 
The army's men almost felt betrayed. They could 
make no progress towards a "Sled." The tensions 
between the SIGINT agencies and IBM were 
quite evident by early 1947. Even OP-20-G had 
become alienated. It had become tired of having 
to bend to the whims of a single supplier and was 
willing to spend extra monies on machines, train- 
ing, and support to gain more bargaining power. 
It was giving very serious consideration to setting 
up a processing center based on the products of 
IBM's competitor, Remington-Rand. 

£ES-) The army's machine group was certainly 
losing patience. It helped draft a protest to the 
"brass" about IBM's lack of cooperation, 51 and 
some rather direct words reached Tom Watson. 
As a result, there was more cooperation. In 1947 
IBM's management made sure that ex-members 
of the agencies who had returned to the corpora- 
tion were assigned as liaison officers. 

(8) Men such as James Green and Stephen 
Dunwell began to do much to restore harmonious 
relations. They arranged little favors such as hav- 
ing IBM replace the frequently wornout parts of 
the SIS keypunches at no cost to the government. 
The cryptoagencies were the only ones at the time 
to do extensive binary punching. That wore out 
the punches and die blocks in weeks, rather than 
in the years that were typical in business data 


Page 215 


@d Green and Dunwell helped keep "F" 
informed of new technological options. They 
made sure that "G" and the SIS were the first to 
know about such important IBM advances as its 
604 electronic multiplier and its very hush-hush 
line of new tabulators. They did more than pass 
information from IBM to the codebreakers; they 
began to act as advocates for SIS and "G." 52 They 
met with IBM management and argued that the 
company would benefit from the Sled research. 53 

(TS//SI) But it was not until the summer of 
1948 that IBM agreed to take on the extensive 
Sled project. 54 Then it was not until that fall that 
the first specific designs began to receive 

(T 8 //SI) Meanwhile, the frustrations at the 
SIS had grown to such a level and the need for a 
Comparator for baudot problems became so 
pressing that the in-house engineers began 
designing an emergency version of a Sled. At first 
called a "tape-comparer," the machine emerged 
in 1948 as a rather crude jury-rigged machine. 

But it worked and it evolved step by step into the 
Connie Comparators of the 1950s. They never 
became as flexible as Sled; but they were seen by 
their sponsors as general-purpose comparators. 55 

(TS//SI) While the Connie precursor was 
being constructed, IBM acquired a secure build- 
ing in Vestal, New York, to work on Sled. IBM 
sent some of its best men to help with the project, 
including a future company president, B. O. 
Evans. 56 

( TS//SI) Unfortunately, just then, long-term 
goals had to be set aside, at least for a time. A 
cryptoemergency arose. In response, a very 
stripped-down version of Sled, with the appropri- 
ate name, Skate, was hurriedly produced and 
rushed to W ashington in 1949 to try to unravel an 
intransigent j system. 57 After a relatively 

long shakedown cruise, Skate was put to work as 
a primitive electronic version of a numeric-only 
Slide-Run machine. It was soon followed by a 
more advanced copy, which cost twice as much as 
the first, over $500,000. 58 

Page 216 

TS) Sled 


P.L. 86-36 
EO 3.3(h)(2) 

TOPocorcnwoo M imv/ncL to uoa, auo, oa n , odr, and N2u/x i 

£B//0 f r The Skates were electronic advances, 
but costly ones in terms of dollars and manpower 
diverted from the Sled ideal. That upset one of the 
major figures in the agency who supported the 
Sled architecture, Albert Highley. He knew from 
first-hand experience of the need for a ubiquitous 
device on the machine room floor. His belief in a 
quickly convertible architecture was perhaps 
reinforced by the Skate experience: By the time 
the machine became fully operational, the origi- 
nal target had disappeared. 

fTST Highley became worried that Sled would 
never be turned into hardware. As a result, he and 
his associate Ray Bowman began to apply new 
pressures on the company. Sled was finally born, 
but that was eight years after Leo Rosen had put 
forward the general outlines of such a machine 
and a year after the SIS and OP-20-G had been 
merged into the new organization, the National 
Security Agency. 59 

(TB// 8 I) What finally arrived in Washington 
in the first half of 1953 were two copies of a cus- 
tom-made machine whose basic design stood, for 
more than two decades, as a tempting alternative 
to the general-purpose computers. The Sleds did 
not achieve all that had been hoped for in 1945 
when the "reconfigurable" design had first 
appeared, and they were more expensive than 
thought. But they were impressive. 

(T0//9I) The two copies cost a third more 
than Madame X,but that was not much more 
than the previous Skate "pilot" models had cost. 60 
And they were inspirations to those who favored 
a special cryptanalytic architecture. 

£P&//$f) The Sleds depended upon high- 
speed electromechanical tabulator equipment 
for their input and output, but they were not ret- 
rogressions. The card reading and punching 
equipment was used because so much of the 
information that was to be processed was already 
in card form and because printers of the time 

were too slow to keep up with the electronics that 
had been developed. 

( TS// S I) The Sleds called upon the best large- 
scale memory technology ofthe era, magnetic 
drums. For the super-fast processes of offsetting 
messages, they used advanced delay-line systems. 

p Ri//S i) The Sleds were built of hardwired 
function "boxes" with very advanced circuitry. 
Although they did not span the full range of 
cryptanalytic functions, those that were included 
gave Sled power over a wide range of cryptanalyt- 
ic problems. 61 Critical to Sled was its type of "pro- 
grams," a combination of plugboards and elec- 
tronic matrices. They allowed instant switching 
and concurrent processing. 

G¥$rff%T) The hardwired functions, the fast 
memories, and the use of plugboard and elec- 
tronic matrix programs were augmented bythe 
ability to have much parallel processing. But with 
or without parallelism, Sled's speed was impres- 
sive. For example, it could make 30,000,000 
comparisons a second if desired In contrast, the 
mid-i940s NCR-Gray Comparators worked in 
the range of hundreds per second. 

fflS//Sfr) Sled could be used as an IC machine, 
a crib-dragger, awired-wheel machine analyzer 
and analog, a statistical threshold tester, and 
much more. And it could be used for alphabetic as 
well as numeric data. 

CES//3 f) One reason for its wide abilities was 
its memory systems. Its magnetic drum held a 
significant amount of data for the time, 48,000 
characters. Its delay lines and special circuits 
which could "precess" (offset) two messages 
made it a very fast comparator and crib-dragger. 
Sled also outdistanced the old Comparators 
because of its thirty-two counters and five accu- 

(T0//0E) Its electronic weighting circuits 
made it a very efficient version of a plaintext 


Page 217 


recognition machine, and its circuits for statisti- 
cal evaluation also helped in the several modes of 
IC analysis. Its "recognition unit" made it a fast 
slide-run machine, and it was a very, very rapid 
"locator." Its two large matrices aided it when it 
was used to decipher systems, including Enigma- 
like ones. 

ffS//S4j Although the Sleds were honored 
because of their slide-run "recognition" abilities, 
a clever engineer could make them perform a 
broad range of functions. One routine made the 
testing of the suspected reuse of key on a major 
system a routine matter. 

( T8// 8 I) For example, 3,000 ten-group por- 
tions of key had been recovered, and it was 
desired to see if any of them had been used on the 
messages that continued to flow in. To do that, all 
the groups had to be applied to the messages and 
the resulting text checked to see which, if any, of 
the keys produced a significant percentage of 
known code groups. Sled was able to test the 
3,000 suspected keys against ten cipher groups in 
just fourteen seconds. 6z 

CTS//SI) "Programming" Sled was an art, 
with the programs looking more like engineering 
timing diagrams than the instructions for a digi- 
tal computer. 63 Despite that, Sled gained so much 
loyalty that the first ones were cloned in a super- 
fast transistor version by the late 1950s. As we 
will see, a grand elaboration was proposed under 
the mid-1950s NSA Farmer program. 64 

(U) Faith and an Institution; the Chance to 

Begin an ERA 

GES//M) OP-20-G had its "reconfigurable" 
machine working some four years before the 
Sleds. 55 The reason for the earlier appearance 
was not because of more engineering genius with- 
in the navy; it was because of different postwar 
institutional arrangements. 

(U) As the war was winding down, OP-20-G 
and the SIS knew they would be stripped of men 
and resources. The situation looked bleak. 
Rosen's "F" branch was in jeopardy, and 
Engstrom's "M" and the NCML faced extinction. 
The Bureau of Ships showed signs of tiring of the 
near autonomous NCML, and "M" had its own 
special problems because of navy personnel rules. 
Holding onto its many exceptional scientists and 
engineers was an especially difficult and pressing 
problem. Without them, little progress could be 
made on methods or machines. There were no 
cryptanalytic think tanks, and all the private com- 
puting machine contractors made no effort to 
hide that they were tired of government work. As 
threatening, in 1945 there was no electronic com- 
puter industry, and there was little indication that 
one would emerge. 

(U) Friedman lobbied the army to maintain as 
many civilian slots as possible, and he tried to cre- 
ate a joint machine development center with OP- 
20-G, but he had to settle for a small group in the 
SIS that could direct and oversee established con- 
tractors. 66 Wenger sought much more. Rejecting 
the suggestions for ajoint army-navy program, 
but later being forced to accept ajoint board that 
sought to coordinate programs and targets, he 
began an independent search for a practical solu- 
tion. 67 

(U) After some initial failures, things began to 
fall into place. The secretary ofthe navy took 
great pride in OP-20-G's achievements, and the 
Chief of Naval Operations had become an ally. 68 
The Office ofthe Chief of Naval Operations 
helped overcome any objections from the bureau, 
and the NCML's life was extended, at least for a 
time. Some postwar funding seemed more than a 

£TS//gf ) Wenger formed anin-house RAM 
panel to take advantage of that and to develop the 
technical arguments he would need to fend off 
any major threats to his automation program. 69 

Page 218 



(U) In late summer 1945 a $500,000, one- 
year development contract was awarded to 
NCML-NCR. It included funds to work on a new 
general-purpose Comparator. Wenger under- 
scored the point that the United States could 
never again expect to have the time to make and 
correct fundamental mistakes as it had during 
World War II. He hammered at two other points: 
The traditional division between operational and 
bureau powers would ill serve a modern navy, 
and only a continuation of something like the 
cooperative relations between "M" group, NCR, 
and NCML could save naval cryptanalysis. 70 

(U) He was given assurances that OP-20-G 
would be allowed its own program and was 
told that navy money would be made available 
for continuous machine development. Then 
Wenger received the wonderful news of the estab- 
lishment of Monogram, a long-term program to 
continue upgrading communications intelligence 
equipment and methods. Hooper's mid-i930s 
plan for naval communications and for linking 
science to the navy appeared to have finally been 

(U) Under project Monogram, every relevant 
research project was placed within one integrated 
program. Radio research, the mathematics of 
cryptanalysis, and even electronic explorations 
relevant to the gathering and analysis of signals 
were to be subject to its generosity. Millions of 
dollars, it was pledged, would be allocated for 
both research and advanced development proj- 


(U) More than money was promised. There 
was a strong hint of autonomy for "G." It would be 
allowed to direct its own work, free from the 
Office ofNaval Research, the Naval Research 
Laboratory, the naval electronics laboratories 
and, to a very great degree, the Bureau of Ships. 
Although the other navy agencies continued the 
battle to control "G's" "turf," the naval Rapid 
Machine program had a future. 72 

(U) In late 1944 Wenger put Howard 
Engstrom, Ralph Meader, John Howard and 
another of the bright navy engineers, Bill Norris, 
to work on Hooper's suggestions. 73 They pro- 
posed what they thought was a way to perma- 
nently link science and innovation to the navy. It 
was anew version of Hooper's post- World War I 
RCA. In 1945 Wenger's men recommended creat- 
ing the private, for-profit, National Electronics 
Laboratory. The company was to be staffed by the 
talented men from OP-20-G and the other 
advanced science agencies in the navy. 74 

(U) Wenger approved the idea, envisioning a 
firm that would devote itself to navy communica- 
tions problems, 75 ranging from mathematical 
cryptanalysis to the physics of radio. 

(U) The navy's legal experts gave the green 
light to"M's" officers, such as Engstrom and 
Norris, having an interest in the private company. 
Most ofthe "M" engineering team, including 
Howard, Coombs, and Steinhardt agreed to join, 
but those who had been IBM employees decided 
to return to their old company. Joe Desch and his 
men also opted to stay with their firm, NCR. 76 

(U) Soon, however, everything seemed to be 
falling apart. America's old scientific organiza- 
tions rejected them. Rockefeller Foundation also 
thought America had enough research institu- 
tions. A sponsor could not be found, and the situ- 
ation became critical. At the end of 1945 Wenger 
had his new research agenda and had promises of 
contracts, but he had no idea of where to find the 
men to build a full electronic Super Bombe, a new 
version of Mike, his grand "reconfigurable" 
Comparator, or even a viable punch for the old 
Comparators and the Copperheads. 77 

(U) A savior, at least a minimal version of one, 
finally appeared and turned Wenger's failing 
dream into the new company, Engineering 
Research Associates. But even the investment 
banker, entrepreneur, and old friend ofthe navy, 
John Parker, could not piece together truly ade- 


Page 219 

EG. 3.3(h) (2) 
P.L. 86-36 


quate funding. He could not even locate the new 
company hear OP-20-G. The proposed research 
arm of "G" had to move to Minnesota. ' 8 Coming 
in contact with the Engstrom-Norris group 
through mutual friends in the military, 79 Parker 
was persuaded that a private version of the 
NCML-NCR could succeed. He agreed to gather 
minimal financing, to help with business matters, 
and to set the new company up in his old factory 
in St. Paul, Minnesota. 

(U) ERA immediately gained the navy's 
approval, and it immediately won OP-20-G's 
big research contract. 80 In return "G" expected 
ERA to be a "captive" of the navy. 

(U) It also won an important friend, OP-20's 
Louis Tordella. One of the young officers who 
decided to stay in"G" after the war, Tordella 
would become one of NSA's most dynamic lead- 
ers. In 1946 he began supervising the ERA con- 
tracts and acted as a general liaison with the com- 
pany. Perhaps because of his interaction with the 
ERA engineers and mathematicians, Tordella 
became one of the future NSA's most energetic 
supporters of high technology as well as one of 
the most influential figures in the history of 
American intelligence gathering. 81 

(U) A Bright Hope for Hooper's Dreams 

OfSffST) By mid-1946 ERA had abroad con- 
tract with "G," one that gave it the freedom Bush 
had sought in the 1930s. Its men were happy with 
competitive salaries, stock in the company, and 
the chance to do cutting edge work in computers, 
communications, and operations analysis. There 
were indications that ERA might also become a 
think tank and a center for advanced mathemati- 
cal research.Those efforts, led by C. B. Tompkins, 
were coordinated with the research of several of 
"G's" alumni who had returned to teach at such 
prestigious institutions as Harvard and the 
University of Illinois. They di d contract work on 
topics such as l l and the behavior of 

binary systems. 82 Wenger had a small but effec- 
tive cadre within OP-20-G to manage his technol- 
ogy program. Joseph Eachus, Howard 
Campaigne, and James T. Pendergrass were top- 
flight young scientists who appreciated the role of 
mathematics and computers in cryptanalysis. 
They helped Wenger set up a board to coordinate 
the RAM program with the needs of the cryptan- 
alysts. 83 

(U) Most importantly, ERA was launched on 
the mission of creating a multipurpose cryptana- 




-,' ijflflj 


(U) Icuis Torctella 

(U) Joe Eachus 

Page 220 



(U) Golcberg 

lytic computer. 84 But ERA did not start its career 
with a leap into fully digital electronic computing. 
It began with an attempt to build the navy's 
version of Sled. 

(U) The Grand Machine of Its Time, the New 

(U) The hopes for a single grand cryptanalyt- 
ic machine had been boiling up at OP-20-G since 
1944. But it took more than a year after Japan's 
surrender before the outlines of the machine 
called Goldberg got the financial nod from the 
Bureau of Ships. 85 Joe Eachus had explored pos- 
sible technologies and sketched ideas which he 
passed onto his old friends who had joined 
ERA. 86 All types of memory media, including 
microfilm, were investigated, as were new tubes 
and circuit designs. 

(U) At ERA the "reconfigurable" general-pur- 
pose Comparator Goldberg began as more of a 
research than a development project. Many of 
Goldberg's components were in advanced stages 
within its first year, but the machine was not 
delivered until late 1949, more than a year behind 
its production schedule, and two and one-half 

behind the hopes of its original planners. 87 Even 
more time was needed to smooth over its opera- 
tions. ERA was not finished with the machine 
until 1951. m Goldberg ended up much more of 
a special-purpose machine than had been intend- 
ed. It did not even become a fully reconfigurable 
computer. Operational needs pushed it to becom- 
ing an elaboration on the early Comparators, but 
one targeted at the new teletype-encryption 

(U) Goldberg took photoelectric sensing and 
paper tape scanning to new technical heights. 
Very fast tape drives were completed by 1947, 
allowing as many as four tapes to be run on top 
of each other. The drives ran the tapes at more 
than six times the speed of the older devices and 
were able to offset the tapes for IC testing without 
slowing the machine. Avery complex and precise 
scanner was developed which included the photo- 
cells and circuits to sense each of the seven data 
and three control positions in each row on a 


(U) Goldberg was also an example of how 
much electronics had matured since 1945 and 
how the emergence of new components could 
undermine investments in the development of 


Page 221 


early technologies. Almost as soon as they were 
developed, the new tape systems were abandoned 
in favor of a series of emerging technologies. 

CE3//SI) The first temptation that pulled 
attention away from tape or film was electrostatic 
storage. For a time it was thought that Goldberg 
was to have special television-like tubes and thus, 
a "random memory." 90 But when RCA and others 
were unable to make such systems operational, a 
"second best" technology was selected for 
Goldberg, the magnetic drum. 

(U) Goldberg was treated to the slower but 
more tractable magnetic drum memory. The sys- 
tem included delay lines and sophisticated cir- 
cuitry that allowed the tracks of information on 
the drums to be offset in the same way that Bush's 
earlier Comparators had slid one tape over anoth- 


(U) Goldberg was given one of the first mag- 
netic drums in the world, and "firsts" always have 
problems. It took several years of effort to make 
the new technology behave. Even in mid-1949 
there were problems with the huge drums on 

£TS//SI) Much more than the drum was inno- 
vative. Goldberg's central cabinets were very 
impressive. They contained more than 7,000 
tubes reflecting the complexity of its digital cir- 
cuits. Going beyond the state ofthe art led 
Goldberg into trouble. Atone point in its con- 
struction the majority ofthe tube sockets in the 
machine had to be replaced. 92 

C5S//SR But once in operation, Goldberg 
was able to perform many different cryptanalytic 
functions through plugboard "programs." That 
helped fulfill some of its designer's hopes that it 
could be a "reconfigurable" machine. 93 

(0//0I) Goldberg could do anything that Mike 
or Copperhead or the old Comparator had done 
during the war and much, much more. It could 

perform so many standard functions that its "spe- 
cialized" architecture seemed more suited to 
cryptanalytic needs than the proposed general- 
purpose programmed computers. 

(S//8f) It performed frequency counts, IC 
tests, round robin searches, crib dragging, wheel 
stripping, roughness tests and, among other func- 
tions, weighted calculations. It had an advanced 
translation system for baud signals and an elec- 
tronic 36 x 36 matrix that imitated, if desired, a 
wired wheel. 9 * It also had a sophisticated thresh- 
old circuit that eliminated the "always print" fea- 
ture of Bush's earlier machine, thus saving much 
run time and analyst's attention. 

(St And Goldberg was fast, although not as 
rapid as the later Sled. When Goldberg was in top 
shape, it made 20,000 serial comparisons a sec- 
ond. That was 250 times the rate of Bush's 70mm 
Comparator. 9S 

( j 8//3 ' Its thirty-six decimal (ring) counting 
circuits allowed deep statistical analysis. In addi- 
tion, it had banks of rectifiers for short-term fast 
memory, which aided the special circuits used to 
calculate the IC statistics. 

63) The Goldberg work led to advances in the 
design and use of magnetic technology despite its 
drums sometimes being taken off to serve the 
emergency needs. Besides its contributions to the 
mechanics and electronics of drum memory, 
Goldberg incorporated a unique way of using its 
drums. They were used as "buffers." To speed 
processing, while one drum was providing data 
for calculations, the other was loaded with data 
from the tapes. 96 

(T3//3I) Although it was late incoming, 
Goldberg was the state-of-the-art "reconfig- 
urable" Comparator. Contemporaries thought the 
more than 8250,000 spent on it was a very wise 
investment, worthwhile enough to think of 

Page 222 



replacing the slow and sometimes troublesome 
drums with a massive electronic memory. 97 

(U//FQUO) However, because of its long- 
delayed and sometimes painful delivery, it 
seemed best not to build any more of the "G" 
designed general-purpose "comparators." The 
contract for a second machine was canceled. The 
progress on the general-purpose computer, Atlas, 
and the old SIS group's faith in the Skate-Sled 
multipurpose machine project at IBM indicated 
that the Comparators had outstayed their wel- 

CB 0//Or) However, the engineers at ERA and 
NSA learned a great deal from its development 
and its perhaps four years of service. They trans- 
ferred much of its technology to other machines, 
including along line of limited comparator-like 
special-purpose devices that were begun in the 
late 1940s and early 1950s. The tape-based 
Robins and Connies and the ambitious delay-line 
Vivians and Delias took much from Goldberg." 

(U) Meanwhile, a Last Chance for Microfilm 

(TS// S I) Although the engineers at ERA had 
decided microfilm was inappropriate for 
Goldberg, there were many in "G" and the SIS 
who continued to have faith in the future of 
Bush's solution to the mass memory problem. 
While the navy had some engineers who saw Icky 
and Hypo as just the beginnings of a major post- 
war microfilm program, it was an army group 
that had the grandest postwar visions for film 
RAMs. Encouraged by the arrival of its microfilm 
plus electronic-counter machine in late 1944 and 
the about-to-be-completed 5202 Comparator for 
the Tunny problem, "F" wanted to order a whole 
series of machines from Eastman. Each would 
perform one of the major cryptanalytic func- 
tions. 100 The ideas were attractive because film 
continued to be a much higher volume memory 
than magnetic devices. Drive speeds did not erase 
the difference. In 1949 microfilm data could be 

read in at five times the speed of data on magnet- 
ic drums. 1C1 

Cny/O fr) The postwar SIS was not allowed to 
launch the ambitious film RAM program, and it 
was unable to have Eastman-Kodak serve as an 
ongoing resource. But the SIS's film advocates did 
begin a project to do the necessary research and 
build an upgraded version of the ambitious 5202. 
Shortly after the war it hired several small elec- 
tronics companies to explore all the possibilities. 
One, Hogan Laboratories, had a promising 
design. A contract was let, and the SIS would have 
a very advanced film Comparator before the end 
of the decade. 102 

(TG//0f 7 Those who continued to see the 
future of OP-20-G in terms of microfilm received 
some funding. With a rather handsome allocation 
in hand, they eventually convinced Eastman to do 
more than complete World War II's Ambers. In 
1947 Eastman accepted a contract to begin to 
explore possibilities for a new Hypo and a new 
Icky. It would take quite a while for Eastman to 
deliver the new versions, but in 1947 film again 
seemed to have at least the possibility of a rebirth 
at "G." 103 There was even some thought of having 
the Eastman group under Tyler build alight- 
based bombe and a new electronic rotor bank. As 
promising was an exploration of a grand idea for 
a huge new comparator using large photographic 
plates or drums with as many as 10,000 tiny 
holes per square inch. As the plates could be 
aligned over each other in one-tenth of a second, 
a light-based machine to attack the new teletype 
encryption machines seemed within reach. 104 

(U) Finally, the Electronic Bombe 

(T0//6I ) In 1947 there was even more that 
gave indications that the cryptanalytic and tech- 
nological triumphs of World War II would con 

against the high-level systemsj_ 


Jsystems allowing the engineers at "G" and 

tinue. Most important was the growing success 

against the high-level systems ] 

Enough had been learned about several of the 



Page 223 

P.L. 86-36 
EO 3.3(h) (2) 

tv-i,. 8 6-36 
E0 (2) 

top 3eeReTOcoM i NT/m e L to usa, aus, caw, qbr, and nzuxi 


new vers ions of the Purple 
relay devices (now 

the SIS to bull 

machines. The 
especially efficient because of a new. IBM relay 
technology) were given the names of less intru- 
sive colors, such as Tan and Pink. 


( TO//0I) There was also a series of e 

mechanical analytic machines for the 1 

problem. The small machines built by the in- 
hbuse engineers were also given "softer" names 
\than those used for similar machines of World 
War II. The Stork was one of many helpful 
devices used directly by the cryptanalysts. 105 


' (TC//8I) Achievements against the 
get went beyond tabletop relay boxes. Perhaps 
the first operating machine to use a magnetic 
drum was constructed as a crib-dragger to attack 
the very\ very im portant on-line encryption 
device th e] used for much of their top ech- 

elon traffic. Beginning work in early 1948, and 
using magnetic drums taken from Goldbeig, ERA 
finished the first of several Demons in October. 
Although the Demons had many relays and plug- 
boards, they had electronic components and the 
circuitry needed to search for high-frequency 
clear groups. Follow-on models were flexible 
enough to be used against several targets. 

0TS//BI) The idea behind the Demon attack 
was clever. A large number of cribs were applied 
to one message of a pair determined to have to 
been produced by the. same key through IC and 
similar analyses. The derived key from the first 
crib-plain match was applied to the other mes- 
sage. Then, to See if true key had been found, the 
result was looked up in the memory, which con- 
tained known high-frequency groups. 106 

(TS//SI) More thrilling for the machine 
builders was the chance to at last construct a full 
electronic Bombe. The explorations ofthe elec- 
tronic matrices and the growing knowle dge ofthe 
rotor-based, on-lincj_ "l eneryptor 

ect. 107 It was one ofERA's great challenges 
because Hiawatha might cost $1,000,000 and 
call for 40,000 tubes and because ERA's engi- 
neers hoped they could construct and design it so 
that it could attack more than one teletype 

£PS//0B Hiawatha was only a be ginning. To 

cover the entire spectrum o j devices, "G" 

led to the appropriately named Hiawatha proj- 

began the Ophis project. Its first goal was anoth- 
er electronic rotor, one for an attack on the mys- 
terious Albatross machine. Albatross was thought 
to be like Germany's wired-rotor Green Enigma 
ofWorld War II. The SIS and "G" hoped that 
Ophis' long-term result would be a general wired- 
wheel Bombe that would be more powerful than 
Hiawatha. 108 

(TS// - 6 f ) There were even greater and more 
exciting engineering challenges. By the time 
Hiawatha was conceived, both ofthe American 
cryptanalytic agencies were joining the race to 
complete the first modern "von Neumann" type 
of computer. 

(U) Notes 

l- ffS//» ) NSA CCH Series XII Z,"ANCIB 
Minutes, abstract of," 1955. This also states that the 
U.S. was much more technically advanced than the 
British. The only edge the British had, it stated, was in 
"collateral" information. ( TS//89 NSA CCH Series XII 
Z, draft copies of Michael L. Peterson, 'The Bourbon 
Problem," indicates that early British successes against 
Russian cipher machines and their ability to intercept 
and process non-Morse transmissions were critical to 
convincing die American agencies to extend the World 
War II cooperation and formalize it in the BRUSA 
agreement. (¥8/7*1) NSA CCH Series XII Zand AHA 
Series IVE.1.1, George Howe, "Historical Study of 
COMINT Production, 1046-1949," April 1957. 

2. ( TS//SI) NSA CCH Series XII Z, "Procurement 
ofGeheimeschreiber Equipment from British," J.N. 
Wenger, OP-20-G, 14 August 1945- 

3.4K-) NSA AHA ACC 78098, "Monogram Report, 
Part IV, Field Research," is useful on multiplex needs. 

Page 224 



P.L. 86-36 

4. (¥S) On IBM's reluctance to take on any work 
after the war and even into 1947, (¥9} NSA CCH Series 
XII Z, file folder, "Monogram and RAM Panel 
Reports," 1945-1949. 

5. (U) By the 1950s the physics community would 
seek an alternative to the classic computer architecture 
as the demand for increased computing power escalat- 
ed. But in the 1940s the codebreakers took the lead in 
seeking a different type of electronic computer. 

6.(¥S-) NSA CCH Series XII MPRO, Box 1, 
"Machines in the Service of Cryptanalysis," 28 
September 1954. 

7. (TS//SI) NSA CCH Series XII Zand AHA 
Series IVE.1.1, George Howe, "Historical Study of 
COMINT Production, 1946-1949," April 1957. 

8. (T0//0IJ NSA CCH Series XII Zand AHA 
Series IVE.1.1, George Howe, "Historical Study of 
COMINT Production, 1946-1949," April 1957, 173. 

9. f¥S) The hopes of the army cryptanalysts were 
expressed in a long memorandum from Frank Rowlett 
to the commander at Arlington Hall in mid-1945, 
(3«) NSA AHA ACC 26373, Frank B. Rowlett, "RAM in 
Future Cryptanalysis," 3 May 1945. Note, however, 
that his vision was less far reaching than those 
expressed inWenger's plans for RAMs atOP-20-G. 
For example, see (28) NSA CCH Series XII Z, CNO 
to Chief Bureau of Ships, "Communications 
Intelligence: Research and Development," 21 
December 1945, and, Wenger toEachus, "Analytical 
Machinery Panel," 31 October 1946. 

10. (¥9//ST) NSA CCH Series XII Zand AHA 
Series IVE.1.1, George Howe, "Historical Study of 
COMINT Production, 1946-1949," April 1957. 

11. (TS//SI) CFS3 NSA CCH Series VI.1.8, 
"Martini," circa 1947- Onlnitram, (¥9) NSA CCH 
Series XII Z,J. F.Beatty, "Martini (Longfellow)," 
OP-20, 1947. The old Japanese analogs, such as 
Python, contributed their parts to the anti-Soviet 
cause. See also the army's Tan analog of the 
Longfellow machine. Exciting sources on the progress 
against one of the Soviet's most sophisticated and 
important cipher machines are found in ( TS//SI) NSA 
CCH Series XII Z, "Longfellow, History of," N-31 to 
20-L, June 1948. 

12. e rS// S i ) (TS) NSA CCH Series XII Z, H. H. 
Campaigne, "Summary of War Diary September 

1946," 7 October 1946. ( TS//SI) NSA CCH Series XII 
Z, H. H. Campaigne to 20-34 L, "Longfellow, History 
of." June 1948. (W€) NSA CCH Series V.I.1.20, 
"Longfellow, Machine Breaking, 1947-" ( TS//SI ) 
_NSA CCH Series XII Z,"0P-20-G, /'History of Navy- 
Attack on Longfellow," 14 December 1949. 

13. (TS//ST) On the frustrations caused by the 
Soviet I | (T0//0I) NSA CCH 
Series XII Z, "File Kept by Dr. Campaigne on Ram 
Panel Meetings." One large set of comparators, the 
Robins, was built just for the problem but achieved lit- 
tle. (- TS//S *) NSA CCH Series XII Z,AFSA-2l 
"Summary of the Early Operation of the Robin 
Machinery," 19 May 1951. 

14. CES // Sfr ) When the world situation began 
to unravel in the late 1940s, not enough was known 
about some of the more important Soviet machines to 
allow a fast technological solution such as the Bombe 
program of World War II. The United States did not 
know enough about the wired-wheel Albatross 
machine, for example, to allow the construction of a 
massive analog-analytic machine like the Bombe. 
(TS#») NSA AHA ACC 18669, AFSA-02, "Request 
for establishment of Comparator Project (Albatross)," 
6 June 1950. 

15. (IS#SI) Oliver R. Kirby, "The Origins of the 
Soviet Problem: A Personal View," Cryptologic 
Quarterly , Vol. 11 #4 (Winter 1992), 51-58, gives an 
insight into the SIS' fears about being ordered not to 
pursue the Soviet problem. (TS//0I) The series of his- 
tories by CCH historian Michael L Peterson published 
in Cryptologic Quarterly, 1994-95, shows the navy 
was working under the same fears. 

16. CES#8*) NSA CCH Series XII Z, 
"Communications Supplementary Activities, RAMP 
Report II," 21 December 1948, provides a summary of 
the devices planned at the end of the war. 

17. C ES// 8*) OP-20-G invested much in the inno- 
vative Hecate machine, which was in the planning 
stage by the end of the war. It is described below as is 
a more general -purpose device, Al catraz, which was at 
first targeted at | ) . (TQ//3I) NSA CCH 
Series XII Z,AFSA-35iB, "The Use of Hecate in 


] October 1950. 

18. ( TS//0I) On Hecate, (SS} NSA CCH Series XII 
MPRO, Box 1, "Machines in the Service of 

TOP IFrOFTIimmUTIIBPl m MCA ahc n«u ^p D »»n i |7 | fft 

Page 225 

EO 3.3(h) (2) 
P.L. 86-36 


Cryptanalysis," 28 September 1954. pE fl//8i ) NSA 
CCH Series XII Z, "Mechanization in Support of 
Comint, Phase II," circa 1955. @S) NSA CCH Series 
XII Z, "General and Special Purpose Computers; A 
Historical Look and Some Lessons Learned," 23 May 
198 (Hogan). And there seemed no need to rush its 
development and construction. 

19. (TS//SI) NSA CCH Series XII Z, 
"Communications Supplementary Activities, RAMP 
Report II," 21 December 1948, states that 32 Bombes, 
2 Grandads and 4 Duennas were still available as of 
1-31-49. The Autoscritcher was retired in 1945. 

20. (U) A useful overview of the Hagelin enterprise 
and influence is Boris C. Hagelin, David Kahn (ed.), 
'The Story of the Hagelin Cryptos," Cryptologia, 
XVII # 3 (July 1994), 204-242. 

22. (¥8) NSA CCH Series XII Z,L. R. Steinhardt, 
"Digraph Counter, Improved, Conference On," 11 July 


23. CTO//0I) Oliver R. Kirby, "The Origins of the 
Soviet Problem; A Personal View," Cryptologic 
Quarterly , Vol. 11 #4 (Winter 1992), 51-58. Note that 
in 1945 and 1946 even the president was unsure of the 
relationship between the United States and the Soviets 
and argued against reading their messages 

24. (TS#Sfl f»} NSA CCH Series XII Z, L. R. 
Steinhardt, "Digraph Counter, Improved, Conference 
On," 11 July 1945, (TS//SI) NSA CCH Series XII Z, 
"Office of Computers, List of Computer s," nd. 
{XS)> NSA CCH Series XII Z, The System I 

(Alcatraz)," AFSA-35lBm circa 1950. (S) NSA CCH 
Series XII Z,"ERA Task #7 Alcatraz," circa 1949. 
4IS} NSA CCH Series XII Z, file folder, "Monogram 
and RAM Panel Reports," 1945-1949- There is some 
confusion in the records over the production of a Baby 
Alcatraz. The best judgment seems to be that when the 

proposed size of machine was reduced, it was called, 
by some, "Baby," and that only one machine was built. 

25. ( C// 6 9 NSA CCH Series XII Z, Sam Snyder, 
"Draft Document, Pre-Computer Machines in Support 
of Cryptanalysis," circa 2 February 1978. 

26. {$) NSA CCH Series XI K, Sam Snyder, Box 12, 
"Analytic Machinery Principles," September, 1949, 37. 
( TB//0I) NSA CCH Series XII Z, LeRoy H. Wheatley. 
"Cryptanalytic Machines in NSA" 30 May 1953. 
(U) Charles J.Bashe etal., IBM's Early Computers, 
(Cambridge: The MIT Press, 1986), 464. 

27. (X8//SI) (ft NSA CCH Series XI K, Sam 
Snyder, Box 12, "Analytic Machinery Principles," 
September, 1949, 37. ( TG//0r) NSA CCH Series XII 
Z, LeRoy H. Wheatley, "Cryptanalytic Machines in 
NSA" 30 May 1953. (U) Charles J. Bashe et al., IBM's 
Early Computers (Cambridge: The MIT Press, 1986), 
46. NSA's precursors acquired at least the next and 
more powerful version of the 603, the 604, and put it 
to use with tabulators stripping additives. 

28. CS#SI} (U)) James W.Cortada, Historical 
Dictionary of Data Processing: Technology (New 
York: Greenwood Press, 1987), 366. See also the 
unpublished work on the early Remington-Rand 
machines by this author. 

29. (S S//S f j NSA CCH Series XII Z, A M. Gleason, 
"Inversion of Matrices with O'Malley," 1948. 

30. ( TS//SI) NS A CCH Series XII Z,"Use of 
HECATE 1 1 Message Placement," October 

1950, and (T8//8I) NSA CCH Series XII Z,"File Kept 
by Dr. Campaigne on Ram Panel Meetings." 

31. ffg/ZS^NSA CCH Series XII Z, NSA "MPRO 
Technical Reports," circa 1956. 

32. (W//89 NSA CCH Series XII Z, A, M. Gleason, 
"Inversion of Matrices with O'Malley," 1948. 

33. (m NSA CCH Series XII MPRO, Box 1, 
"Machines in the Service of Cryptanalysis," 28 
September 1954. 

34. (WW) (Laconici Nocon ) NSA CCH Series 
XII Z, Glenn F.Stahly, "Fifty Yeate of Mathematical 
Cryptanalysis," August 1985. (TS//SI) NSA CCH 
Series XII Z, "Office of Computers, List of Computers," 

35- (U) NSA CCH Series XII Z, NSA-OH-07-83, 
Oral History Interview with Beverly It; Chall, 2 May 

Page 226 


P.L. 86-36 
EO 3.3(h) (2) 


36. e rS//SI ) NSA CCH Series XII Z, 
"Mechanization in Support ofCOMINT, Phase 
II,"circa 1955- 

37. (¥S) NSA CCH Series XII Z, ERA, Contract 
Number Nobsr-42001, "Preliminary Report and 
Proposal, Task: Project Warlock," 9 June 1948. (S) 
NSA CCH Series XII Z,ERA, "Warlock Progress 
Reports, ERA Task 18," 21 November 1947 to 10 April 


38. (TS//SI) NSA CCH Series XII Z, "Office of 
Computers, List of Computers," nd. 

39. ffS//8ft (Laconicj Nnnnn) NSA CCH Series 
XII Z, Glenn F.Stahly, "Fifty Years of Mathematical 
Cryptanalysis," August 1985. (IS#H) NSA CCH 
Series XII Z, "File Kept by Dr. Campaigne on Ram 
Panel Meetings." f[5i34]- W7W) NSA CCH Series 
XII Z, "Fifty Years of the Soviet Off-line Machine 
Cipher," 10 January 1989- 

40. ( Tfl//0ft NSA CCH Series XII Z, ERA, 
Contract Number Nobsr-42001, "Preliminary Report 
and Proposal, Task: Project Warlock," 9 June 1948. 
Warlock borrowed from the Whirlwind Project at MIT. 
But it seems to have used a unique three-value logic for 
its weighing system to save tubes and processing time. 

41. ( T0//S i) Apparently, the rush of work in World 
War II led to the Americans not gaining enough skills 
to make independent attacks on the Tunny-like 
machines. P&fttt) NSA CCH Series XII Z, S-2733, 
"Longfellow, History of," by Howard Campaigne, June 
1948. (IS//S*)- Oliver R.Kifby, 'The Origins of the 
Soviet Problem: A Personal View," Cryptologic 
Quarterly, Vol. 11 #4 (Winter 1992), 51-58. A list of 
priorities from late 1947, after world events had put 
more pressure on the agencies, still reflect the faith in 
a general-purpose machine. ( - TS//SI) NSA CCH Series 
XI K, Sam Snyder, Box 17, "Long Range Cryptanalytic 
Program for Literal Systems," December 1947. 

42. (TS//SI) The philosophy oP'reconfigurable" 
machines and universal components underwent sever- 
al modifications during the era, but the goal remained 
the same. For a later interpretation, (¥8} NSA CCH 
Series XII X-MPRO, U.S. Cryptanalytic Research and 
Development Committee, "Joint Long Term Program 
for Research and Development in the Field of 
Cryptanalytic Equipment," 21 July 1948. 

43. fFS^&I) NSA AHA 36746, Engineering 
Research Associates, Inc., "Proposal for An Electronic 
Rotor Program," 19 December 1946. On the continued 
search by the navy for high-speed components for the 
matrix, the Leo Project, which came to include the 
exploration of most basic technologies, including 
saturable cores, @) NSA CCH Series XII Z,LEO: 
progress Report, ERA Task #11, 1 September 1947 - 
1 October 1948. On the army's extensive basic 
electronic research during the late 1940s, (TS//SI) 
NSA CCH Series IV.C.2.14, ASA "Annual Reports for 
the Fiscal Year 1949, Vol. II, Research and 
Development Division," Washington, 30 June 1949. 
The amounts invested in such research indicate that 
the SIGINT agencies must have played a significant 
role in pushing the development of tubes and transis- 
tor technology. 

44- ( TS//W ) NSA CCH Series XII Z, 
"Communications Supplementary Activities, RAMP 
Report II," 21 December 1948, 9, shows that Sled 
became a joint army-navy project with the Bureau of 
Ships managing the contract. But the Sled concept 
seems to have originated at the SIS. 

45. (TS//SI) On the army plans, Rosen (T8C) NSA 
CCH Series XII Z, "File Kept by Dr. Campaigne on 
Ram Panel Meetings;" on the navy's (18) NSA CCH 
Series XII Z, "General and Special Purpose 
Computers: A Historical Look and Some Lessons 
Learned," 23 May 1986, (Hogan); and on the navy's 
participation in Sled 4S) NSA CCH Series XII Z, J. H. 
Howard, "Conference on Slid(e)-Run Machine," 
5 January 1946; (IS)- NSA AHA 36746, Engineering 
Research Associates, Inc., "Proposal for An Electronic 
Rotor Program," 19 December 1946; CES//SI) NSA 
CCH Series XII Z, "Joint Long Term Program 
(Old Planning Material, 1948-1949) compiled by 
Doug Hogan. 

46. ( TS//SI ) Leo Rosen outlined his idea of a 
reconfigurable electronic machine in fT_S//Sf) NSA 
CCH Series XII Z,"File Kept by Dr. Campaigne on 
Ram Panel Meetings." 

47.4 TS//Sft NSA CCH Series XII Z, "File Kept by 
Dr. Campaigne on Ram Panel Meetings," Joint 
Research And Development Board Memoranda. 

48. (X S/fOf T NSA CCH Series XH Z, "File Kept by 
Dr. Campaigne on Ram Panel Meetings. 


Page 227 


49- (Sffm NSA CCH Series XII Z, J. H. Howard, 
"Conference on Slid(e)-Run Machine," 5 January 

50. (WtW) NSA CCH Series XII Z,file folder, 
"Monogram and RAM Panel Reports," 1945-1949, 

51. fFS» NSA CCH Series XII Z, file folder, 
"Monogram and RAM Panel Reports," 1945-1949. 

52. (S). On the 604, (S) NSA AHA ACC 8544, 
"Memorandum for Members of RAM Panel, New I. B. 
M. Tabulator," circa 1948-9. 

53- (U) NSA CCH Series XII Z, folder marked, 
Snyder, "Precomputer Comments," circa 1978, 
"Possible Item of Interest." 

54. ff3//39 NSA CCH Series XII Z, 

"Communications Supplementary Activities, RAMP 
Report II," 21 December 1948. 

55. (TS//SI) NSA CCH Series XII Z, James L. 
Sapp, The Analytic Machines," circa 1955. (TS//SI) 
NSA CCH Series XII Z, LeRoy H.Wheatley, 
"Cryptanalytic Machines in NSA" 30 May 1953. 
£5S) NSA CCH Series XII Z, Sam Snyder, draft copy 
of, "Pre-Computer Machines in Support of 
Cryptanalysis," circa February 1978. 

56..0S7W NSA CCH Series XI K Snyder, Box 10, 
10-27-77 Folder. 

57. (ST J.J. Eachus, "SIGMAGE Threshold 
Control," 2 July 1946. & NSA CCH Series XII Z, 
BuShips, "Specifications Sled Navy Models CXOA and 
CXNQ Block Diagrams," 1 October 1948. mffm) 
NSA CCH Series XII Z, "Communications 
Supplementary Activities, RAMP Report II," 
21 December 1948. £ES} NSA AHA ACC 10842, 
Ray L. Bowman," Engineering Diary ,"circa 1945-1950. 

58. (S6//0fl On the Skates (Q NSA CCH Series 
XII Z, Descriptions of NSA Early SPDs and 
Computers, as compiled from various NSA sources, 
and, (C//CI) NSA CCH Series XII Z, Herbert W. 
Worden, "EDP Machine History." Apparently both 

in late 1948. See Michael L. Peterson, (T3//SI ) 
"Beyond Bourbon," 1948, 4- The Skates were flexible 
enough, however, to be used on other problems. 

59. ffS^mff NSA AHA ACC 10842, Ray L. 
Bowman, "Engineering Diary," circa 1945-1950. 

60. (TS//SI) NSA CCH Series XII Z, Samuel S. 
Snyder, "Pre-Computer Machines in Support of 
Cryptanalysis," draft, circa 16 February 1978, IV-25. 
Offi#H) NSA CCH Series XII Z, "Office of Computers, 
List of Computers," nd. 

61. (T0//S1) The Sled was constructed of two large 
interconnected cabinets; the term "boxes" is used for 

62. ffSjr NSA CCH Series XII MPRO, Box 1, 
"Machines in the Service of Cryptanalysis," 28 
September 1954, 10. 

63. Wf$T) NSA CCH Series XII Z, James L. 
Sapp, "The Analytic Machines," circa 1955. 

Technical Reports," circa 1956. The price tag was not 
inconsequential. The transistor version cost approxi- 
mately $2,000,000. 

65. (T8//0I) NSA CCH Series XII Z, "Office of 
Computers, list of Computers," nd. A pilot model of 
the proposed Sled, which was completed to help with 
the Soviet problem, was Skate. It arrived in late 1949; 
the version that was closer to the original grand inten- 
tions, Sled, was delivered in early 1953. 

66. (U) Friedman sought a joint army-navy pro- 
gram, but the navy never accepted the idea. NARA 
RG457, SRMA-011, "Senior Staff Meeting Notes," 
April 3, 1945, "Friedman's joint work suggestion," 174, 
231, 321. Samuel S.Snyder, "Abner: The ASA 
Computer, Part 1: Design," NSA Technical Journal, 25 
(1980): 49. On turf battles among the services, Louis 
Kruh, "Army-Navy Collaboration for Cryptanalysis of 
Enemy Systems," Cryptologia, 16 (1992): 145-164. 

67. (TS//SI) NSA CCH Series XII Z, "Joint Long 
Term Program (Old Planning Material, 1948-1949)" 
compiled by Doug Hogan. 

68. CW-Robert William Love, Jr., The Chiefs of 
Naval Operations (Annapolis: Naval Institute Press, 
1980), 137-192. 

69. (Wf-m) NSA CCH Series XII Z, "File Kept by 
Dr. Campaigne on Ram Panel Meetings." 

70. (U) NSA RAM File: August 21, 1945, 
"Continuation and Development of Communication 
Intelligence"; Part II of Report to J. N. Wenger, Capt. 
USN, "Resume of the Dayton, Ohio Activity During 
World War II," December 21, 1945, CNO toBoShips, 

Page 228 


P.L. 86-36 
EO 3.3(h) (2) 


"Continue to fund NCML"; and March 21, 1946, OP-20- 
G "History of Formation of ERA." 

71. (U) NSA AHA ACC 40731A, SRMN-084, "The 
Evolution of the Navy's Cryptologic Organization." The 
importance and scope of Monogram for advancing inter- 
cept capabilities and mathematics in cryptanalysis is 
reflected in, (IS) NSA 40 AHA ACC 7808, "Monogram 
Report," 29 November 1949. (XS). NSA CCH Series XII Z, 
"Report of the Second Computer Study Group," as in 
NSA Technical Journal, XTX (Winter 1974), 21-61. 

72. £8) (U) NSA RAM File, December 20, 1945, 
"ERA postwar research plan," and July 20, 1946, 
Engstrom: BuShips,"Use Naval laboratories, not ERA" 
ON the Bureau of Electronics attempts to control OP-20- 
G's part in Monogram, f¥9) NSA AHA ACC 7808, 
"Monogram Report," 29 November 1949. 

73. (U) NARA RG457, SRH-267, "History of 
Engineering Research Associates." NSA RAM File, 
September 12, 1947, "Minutes ofOP-20-2 Research 
Committee Meeting." 

74. (U) NARA RG457, SRH-267, "History of 
Engineering Research Associates." NSA RAM File, 
January 2, 1945, Wenger OP-20-G to CNO, "Plan for 
ERA," and August 21, 1945, "Continuation and 
Development of Communication Intelligence [ERA]." 

75. (U) NARA RG457, SRMN-084, "The Evolution 
of the Navy's Cryptologic Organization," 15. 

76. (U) Bright mathematicians and physicists also 
joined the new company. Hagley Museum and Library, 
Accession 2015, Unprocessed ERA Materials, ERA 
Personnel Summaries, circa 1946, and Engstrom to 
Norris, September 11, 1946. The Staff of Engineering 
Research Associates, High-Speed Computing Devices 
(New York: McGraw-Hill, 1950). All departments of the 
navy were concerned about how to continue their 
advanced scientific work. U. S. Naval Administration in 
World War II, War History of the Naval Research 
Laboratory, Guide No. 134, and Harvey M.Sapolsky, 
Science and the Navy: The History of the Office of Naval 
Research (Princeton: Princeton University Press, 1990). 

77. (U) Hagley Museum and Library, Accession 
2015, Unprocessed ERA Materials, Engstrom to Norris, 
September 11, 1946. NSA RAM File, December 20, 1945, 
"ERA post war research plan," and December 21, 1945, 
CNO to BUShips, "Continue to fund NCML." 

78. (U) The Charles Babbage Institute holds many 
informative inter-views with ERA founders. 

79. (U) Important was Nelson Talbott, the powerful 
Dayton business executive. 

80. (U) Charles Babbage Institute, "An Interview 
With James Henry Wakelin, Jr.," OH 104, Conducted 
by Arthur Norberg, February 27, 1986. Hagley Museum 
and Library, Accession 2015, Unprocessed Remington 
Rand / ERA materials, ERA Minute books 1946. NARA 
RG457, SRH-267, "History of Engineering Research 
Associates," 6-7. NSA RAM file: March 8, 1946, John 
Parker to Secretary of the Navy, "Plan for ERA"; 
March 8, 1946, OP-20-G, "List of research projects and 
secret ERA contract of 12-21-45"; and March 21, 1946, 
OP-20-G "History of Formation of ERA" 

81. (U) On the expectations that ERA would be a 
strictly navy firm, (U) NSA AHA ACC 40731A, SRMN- 
084, "The Evolution of the Navy's Cryptologic 

82. (TS//SI ) NSA CCH Series XII Z, 
"Communications Supplementary Activities, RAMP 
Report II," 21 December 1948. A somewhat later but very 
interesting project was concentrated on the mathematics 
of sorting.-^ NSA CCH Series XII Z, ERA A. E. Roberts, 
"An Experiment in the Rearrangement of Data 
(Sweater)," (Sorting, Nomad) 1 May 1950. 

83. PSftSfi NSA CCH Series XII Z, 
"Communications Supplementary Activities, RAMP 
Report II," 21 December 1948. 

84. (U) NSA RAM File, December 20, 1945, "ERA 
postwar research plan." Hagley Museum and Library, 
Accession 1901, Yuter Papers, June 6, 1946 to July 28, 

1946, ERA-NCML on "Orion-Goldberg Project," and 
August 4-8, November 1-9, 1946, ERA reports "Orion- 
Goldberg, binary and analog magnetic recording." 
Hagley Museum and Library, Accession 2015, 
Unprocessed Remington-Rand / ERA materials, August 
17,1946. "ERA salaries." NSA RAM File: August 14, 

1947, Bureau of Ships toERANCML "Task contracts 
causing problems"; June 3, 1946, NCML to ERA, "have 
your work approved"; and July 22, 1946, CNO to 
Secretary of the Navy, "Project Monogram." 

85. CE S//8 3 NSA CCH Series XII Z, "File Kept by Dr. 
Campaigne on Ram Panel Meetings." OP-20-G had 
asked IBM to take on a long-term research contract in 
1945 and again in 1946, but was refused. 


Page 229 


P.L. 86-36 
EO 3.3(h) (2) 

86. (U) Hagley Museum and Library, Accession 
1901, Yuter Papers, September 1946 November 1, 1946 
reports "Orion-Goldberg, binary- and analog magnetic 

87. a&JJ&Pr NSA CCH Series XII Z, 
"Communications Supplementary Activities, RAMP 
Report II," 21 December 1948." 

88. (XS//ST) NSA CCH Series XII Z, "Office of 
Computers, List of Computers," nd. © NSA CCH Series 
XII Z, "Goldberg Progress Reports," 30 December 1947 
through 10 April 1951. (»ef) NSA CCH Series XII Z, 
"File Kept by Dr. Campaigne on Ram Panel Meetings." 

89. (U) Hagley Museum and Library, Accession 
1901, Yuter Papers: Goldberg Reports July-August, 
1946; January i, 1947; and September 22, 1947. 

90. ( IS//flfr- NSA CCH Series XII Z, 
"Communications Supplementary Activities, RAMP 
Report II," 21 December 1948. 

91. £») NSA CCH Series XII Z,file folder, 
"Monogram and RAM Panel Reports," 1945-1949, gives 
the date of the commitment to magnetic drums as early 

9 2. (TS//SI) NSA CCH Series XII Z,ffle folder, 
"Monogram and RAM Panel Reports," 1945-1949. 

93- CES#S0 NSA CCH Series XII Z, "General and 
Special Purpose Computers: A Historical Look and Some 
Lessons Learned," 23 May 1986. 

94. m NSA CCH Series XI K, Sam Snyder, Box 12, 
"Analytic Machinery Principles," September 1949. 

95. (S#SI) NSA CCH Series XII Z, "Symbols with 
their meanings for GOLDBERG programming," nd. On 
the speed of the 70mm Comparator, (**$) NSA CCH 
Series XII MPRO, Box 1, "Machines in the Service of 
Cryptanalysis," 28 September 1954. 

96.Ǥ*-NSA CCH Series XI K, Sam Snyder, Box 12, 
"Analytic Machinery Principles," September 1941 43- 

97. ffS//# r) NSA CCH Series XII Z, "Office 
of Computers, List of Computers," nd. On the drum 
problems,-^ NSA CCH Series XII Z, "Goldberg Progress 
Reports," 30 December 1947 through 10 April 1951. 

98. (U// E0 U6 ) NSA CCH Series XII Z, Sam Snyder, 
"Draft Document, Pre-Computer Machines in Support 
of Cryptanalysis," circa 2 February 1978, IV-15. 

99. fflfl//3 i) (U) Hagley Museum and Library, 
Accession 1901, Yuter Papers, ERA, Goldberg Reports, 
July 1946, September 1946, and January 1,1947. On 

Vivians, CES/#9 NSA CCH Series XII Z, "Office of 

Computers, List of Comp uters," nd. and their use v 

l fffl//03 NSA CCH Series XII 

Z, "Mechanization in Support ofCOMINT, Phase II," 
circa 1955. 

100. gF0//0 i) NSA CCH Series XII Z, "History of the 
Signal Security Agency, Volume Two, The General 
Cryptanalytic Problems." CP8) NSA CCH Series XII Z, 
"The Status of RAM," circa June 1945. {8} NSA AHA 
ACC 26373, SIS, "Technical Paper, RAM," circa June 
1945- 68) NSA AHA ACC 29373, SIS Chief "F' Branch, 
"Request for RAM Equipment," 23 March 1945. &) NSA 
CCH Series XII Z.0P-20-G, "SSA Proposal for 70mm 
Film I.C. Machine," 8 June 1945. 

101. (S)JJSA CCH Series XI K, Sam Snyder, Box 12, 
"Analytic Machinery Principles," September 1949. 

102. tf 3//MF NSA CCH Series IV.C.2.14, ASA, 
"Annual Reports for the Fiscal Year 1949, Vol. II, 
Research and Development Division," Washington, 30 
June 1949. (TS//SD NSA CCH Series XII Z, "Office of 
Computers, List of Computers," nd. 

103. (T3//3I) NSA CCH Series XII Z, file folder, 
"Monogram and RAM Panel Reports," 1945-1949. 
HYPO II was not delivered until 1952. There was also 
a project to explore the possibilities of computer output 
microfilm during the late 1940s. The Eastman contract 
was less than one tenth of what was allocated to ERA, but 
it was appreciable. ffS* NSA CCH Series XII X-MPRO, 
U.S. Cryptanalytic Research and Development 
Committee, "Joint Long Term Program for Research 
and Development in the Field of Cryptanalytic 
Equipment," 21 July 1948. (GffBf) NSA CCH Series XII 
Z,H. H. Campaigne, "Conference About Squinter," 
15 November 1949. (T0//8I ) NSA CCH Series XII Z, 
"Communications Supplementary Activities, RAMP 
Report II," 21 December 1948. 

104. CEB^m NSA CCH Series XII Z, "File Kept by 
Dr. Campaigne on Ram Panel Meetings," and, (0) NSA 
CCH Series XII Z, H.H. Campaigne, "Conference About 
Squinter," 15 November 1949. 

105. ( T0//0I) NSA CCH Series XII Z, 
"Communications Supplementary' Activities, RAM 
Report II," 21 December 1948, 19. 

106. (TS//SI) NSA CCH Series XII Z,ffle folder, 
"Monogram and RAM Panel Reports," 1945-1949, and, 

Page 230 



(S) NSA CCH Series XII Z, ERA, "Demon II Progress 
Reports," 15 July 1948 to April 1951. 

107. 0«7VSTJ NSA CCH Series XII Z, 
"Communications Supplementary Activities, RAMP 
Report II," 21 December 1948," states that Hiawatha was 
begun on March 1, 1948, only to face the disappearance 
of Longfellow from the airwaves on April 11. 

108. ( XS//S 9 NSA CCH Series XII Z, "Longfellow, 
History of," N-31 to 20-L, June 1948. £¥S} NSA AHA 
ACC 8252, OP-20-G, "Communications Intelligence 
Research Plans, 1948," 7 April 1947. QES) NSA CCH 
Series XII Z,file folder, "Monogram and RAM Panel 
Reports," 1945-1949- (TS//SI) NSA CCH Series XII Z, 
"Communications Supplementary Activities, RAMP 
Report II," 21 December 1948. 

TOP oronrn/nn -ninn tomca m.q , rau M q ^ Nn N7| m Rage 231 


This page intentionally left blank 

Page 232 TflP acrpcT»prniiMT»PPl Tniirft Al i rr PAN. ODH. AMD NZUflH 


Chapter 8 
(U) Courage and Chaos: SIGINT and the Computer Revolution 

(U) It Wasn't Safe at the Cutting Edge 

(U) Well before the plans for Goldberg and Sled 
had matured, OP-20-G, joined a bit later by SIS, 
started a great adventure. They became part of 
what many see as one of the most important tech- 
nological revolutions in history. The SIGINT agen- 
cies became active players in the attempt to make a 
technical fantasy come true - to build a universal 
machine. "G" and the SIS, along with several other 
military and civilian agencies, became prime 
movers in the early stages of the computer revolu- 
tion. Establishing that historic foothold was not 
easy for either SIGINT group. 

(U) An idea had emerged and became some- 
what formalized outside of the intelligence commu- 
nity by 1945. It was going to be possible to have a 
high-speed electronic computer that could mimic 
any mathematical or logical process. With a rapidly 
changeable program, it had the potential to be a 
machine for every purpose, from calculation to 
machine control. The key to the machine's flexibili- 
ty was its simplicity. It was to have very, very few 
hardwired functions, perhaps just the four basic 
arithmetic ones, and a few that allowed the move- 
ment of data between the input-output compo- 
nents, memory, and the single central processor. 
That and the organization of the machine around 
the binary system would, it was hoped, make it rel- 
atively inexpensive and allow it to become a mass- 
produced product. With one piece of hardware that 
could be made to imitate any machine through an 
inexpensive and easily changed set of instructions, 
the new computer had a great future. It would 
replace all other calculation and, perhaps, data pro- 
cessing devices. 

(U) The ideas for the universal computer that 
began to take definite shape in England and the 
United States in 1946 were very appealing. As soon 

as they heard of them, mathematicians and engi- 
neers within "G" and the SIS pleaded with their 
superiors to make programmed computers part of 
the SIGINT arsenal. They were persuasive. By 1947 
both agencies had committed themselves to acquir- 
ing general-purpose "computers." 

(U) Neither agency realized what traumas they 
would have to go through to obtain them, however. 
Especially in the case of the SIS, the postwar expe- 
rience was as anxiety-filled as the trials that Hooper 
and Wenger had gone through in the mid- 1930s 
when they sought Vannevar Bush's help. 

(U) Because Wenger had been able to set up a 
semicaptive engineering corporation in 1946, OP- 
20-G had an easier time than the army did. But 
even "G" and its Engineering Research Associates 
had some very tough moments trying to make the 
new computer come to life. 

(U) That had not been foreseen in 1946. After 
learning of the possibilities of the new architecture, 
each agency had expected that outsiders would pro- 
vide all that was needed. That was naive. The SIG- 
INT agencies soon found it necessary to do much, 
much more than they anticipated. Because of the 
chaos that marked the development of the comput- 
er industry in the postwar era, both had to create 
their own machines. 

(U) An Idea Differed 

(U) In 1945 while the ambitious Goldberg's 
technology, if not its architecture, shifted with the 
appearance of technical innovations, and while 
machines like O'Malley were being constructed for 
immediate problems, another and more adventure- 
some project began at "G." 1 


Page 233 


(U) Duenna and the other "electronic" 
machines of the last two years of the war, com- 
bined with the knowledge of what other comput- 
er projects in the nation were attempting, gave 
the "M" group some ideas about a general-pur- 
pose computer. It was to be one much more flex- 
ible than Bush's older Rockefeller Analyzer or 
even his purely electronic Rapid Arithmetic 

(U) When they had a few moments for reflec- 
tion in 1944 and 1945, Engstrom and others in 
"M" speculated about what they could accomplish 
if they could find a large and fast memory, such as 
the vastly improved versions of the delay lines 
they were already experimenting with, to add to 
an electronic processor. While RCA's Jan 
Rachman's new idea for an all-electronic comput- 
er was rejected as almost "screwball," "M's" men 
kept thinking about the future. If a large memory 
with a speed that came close to that of the elec- 
tronic processor could be found, then they 
thought a general-purpose computer was a possi- 

(U) But unless there was a high-speed memo- 
ry, electronic processors would have to remain as 
special-purpose devices. Until the software could 
keep up with the speed of the processor, there was 
little need for electronics. If an electronic com- 
puter depended upon tape readers or the like for 
its directions, it could be no faster than the slow 
mechanical components. 

(U) The input speeds of the best tape and card 
readers of the era were orders less than electron- 
ic processors. That limitation was compounded 
by the serial nature of both technologies. It was 
impractical to ask tape and card systems to back 
up to previous positions and repeat the reading of 
data or "instructions." A universal computer 
needed a memory that could support "go to" com- 
mands because tapes and cards could not fulfill 
that need. 

(TO// 3D The limitations imposed by the 
absence of high-speed memory were one of the 
reasons why the Sled architecture seemed so 
appealing. With special "boxes" hooked together 
through plugboard programs, processing was not 
dependent upon the nonexistent memory. The 
absence of memory was also one of the reasons 
why IBM and other business machines manufac- 
turers confined their postwar electronic offerings 
to limited and special-purpose attachments, such 
as multipliers and dividers that hooked onto tab- 
ulators. 2 

(U) In 1945, any engineer who thought about 
moving further than the Aiken-IBM combination 
of motors, shafts, and tape readers, or the Moore 
School's set ofENIAC special-purpose boxes 
"programmed" through resetting huge electrical 
cables, had to have a great deal of faith. He had to 
believe that some technological hints would soon 
become viable and affordable hardware. There 
were some indications that such dreams might 
come true. But in 1945-6 they were just indica- 

(U) Some thought that delay lines, tubes filled 
with chemicals, could be reengineered to serve as 
memories. The young experts at the University of 
Pennsylvania who were building the ENIAC felt 
they could convince delay lines from radar sets to 
behave well enough to hold programs as well as 
the data needed for immediate processing. That 
was a courageous commitment because those 
"acoustic" delay lines were very temperamental. 
It was very difficult to regulate the timing of the 
pulses that flowed through them. Slight changes 
in ambient temperature caused serious distor- 
tions. Also, it was difficult to monitor the behav- 
ior of the crystals that sensed the data "pulses" at 
each end of the tube. Even when all the technical 
difficulties were eliminated, a fundamental prob- 
lem remained. The tubes could hold only a few 

(U) There were some other memory possibili- 
ties being discussed at the end of the war. One 

Page 234 



was to use a variation of the emerging television 
technology to store and recover "dots" of infor- 
mation on an oscilloscope-like screen. If it could 
be made to work, it would be an ultra-fast memo- 
ry. A computer would not have to "wait" until the 
information it needed cycled past a sensing sta- 
tion. It would run at electronic speeds and would 
allow parallel data transmission. 3 

(U) There were more esoteric ideas for power- 
ful memories, such as RCA's Selectron and the 
use of magnetics, but they were even less ready 
than the other alternatives. 

f6} Although the engineers at OP-20-G knew 
of the technological limits, they could not pass up 
a chance to at least survey universal computer 
options. John Howard formalized some of the 
ideas in a June 1945 memorandum; then, along 
with the "G" mathematician, C. B. Tompkins, 
toured all the East Coast computer projects look- 
ing for more ideas. 4 But little came of their trips. 
"G" was too busy to explore other than cryptana- 
lytic machines. That remained true for several 
months after the war ended. Its workload even 
prevented *G" from sending a representative to 
one of the earliest postwar computer meetings. 

(U) When Howard Engstrom received an 
invitation to participate in a major navy sympo- 
sium on computers, he replied that "G" had done 
little of the type of work that was to be discussed 
and that he and his crew were too busy to attend. 

(U) The urge to explore the possibilities of a 
general-purpose computer continued. But little 
could be achieved. "G" found it difficult to acquire 
connections to the outsiders, especially the aca- 
demics, who seemed to be taking the first major 
steps towards creating the modern computer. 
"G's" old scholarly friend and go-between, 
Vannevar Bush, had stepped back from OP-20-G 
when the war broke out and did not try to 
reestablish the 1930s relationship. That left "G" 
without a prestigious outside scientist who could 

provide the critical endorsements speculative 
projects needed. 

(U) Bush also decided not to return to MIT. 
He remained in Washington, acting as something 
of an academic elder statesman and high-level 
science policy maker until his retirement. Among 
his many contributions, he gave advice on the 
future of science in the military. In addition, Bush 
was frequently called upon to make recommen- 
dations concerning the integration of the nation's 
intelligence services. His role as a science advisor 
to President Eisenhower also played an impor- 
tant part in SIGINT mechanization in the 1950s. 5 

(U) Bush stayed at quite a distance from the 
computer developments of the postwar era. He 
also stayed away from OP-20-G, except for a few 
courtesy visits that Joseph Wenger arranged. One 
reason for Bush's arm's length relationship was a 
very heated argument with the Bureau of Ships 
about the Comparator. Soon after the war the 
bureau decided that it should be protected by 
patents. Bush was sent all the necessary paper- 
work to sign. He did so, but only after the deepest 
protests to the navy about revealing precious 
secrets and about imposing upon him. 6 

(U) OP-20-G had lost another friend. 
Stanford C. Hooper was in semiretirement. He 
was now old and ill, and he had to spend much 
time in Florida. He was acting as a consultant to 
several small electronics firms, including ERA, 
however. He still had the ear of many Washington 
influentials, but he could no longer aggressively 
fight to link OP-20-G, the scientific establish- 
ment, and the large corporations. In fact, he had 
become a bit soured on the corporations and aci- 
demia. He had come to favor small private com- 
panies as the only guarantor of innovation and 

(U) Meanwhile, the other part of OP-20-G's 
old university-computer connection, Bush's 
"boys," had migrated to the "captive" corporation, 
ERA. Howard, Coombs, and Steinhardt were 


Page 235 


keeping up with computer developments, but 
ERA's first contracts and the imperative to devel- 
op a "cryptanalytic" machine kept them too busy 
to act as computer innovators. As a result, their 
1945 general-purpose computer aspirations lan- 
guished until mid-1946. 

(U) Then, "G" developed a new and energetic 
computer champion. At the same time, it found 
someone with great enough scientific status to 
validate its request to acquire something which, 
in the mid- 1940s, seemed more fanciful than 
Bush's 1930s machine. 

(U) Goodbye Dr. Bush, Hello Professor von 

(U) Just as the Goldberg project was launched 
in St. Paul and as Wenger's own research group 
was deciding whether or not to have someone 
build an electronic Super Bombe, one oP'G's" 
mathematicians, James T. Pendergrass, enrolled 
in a summer institute on the programmable, dig- 
ital electronic computer. 7 

(U) His inclusion in the Philadelphia meeting 
was almost an afterthought. Apparently "G" had 
not been asked to send someone until a few weeks 
before the Moore School Lectures began. 
Pendergrass had intended to spend much of the 
summer on vacation, but when his boss, Howard 
Campaigne, called him, he found it impossible to 
refuse the assignment. He rushed to the 
University of Pennsylvania and immediately 
began sending reports to Campaigne. 

(U) Howard Campaigne was one ofthose 
bright young men who had been brought into "G" 
early in the war. Like his friend, Joe Eachus, he 
spent much time in England. 8 And like Eachus he 
became deeply involved with the RAM program. 
Deciding not to go to ERA he became a civilian 
scientist within "G." He helped shape and direct 
"G's" postwar research agenda. By 1946 he was 
one of Joseph Wenger's right-hand men and was 
respected enough to be allowed to act as a repre- 

sentative of "G" to the outside world. That was 
what caused him to attend an important Navy 
Department conference in spring 1946. 

(U) The conference was on the nature of 
large-scale computers. The major address was 
given by the man who would soon equal or exceed 
Vannevar Bush's status in the scientific-political 
realm, John von Neumann. 9 

(U) John von Neumann was perhaps the most 
famous of the new applied mathematicians. He 
had migrated from Europe in the 1930s to join 
the likes of Albert Einstein at America's only true 
research institute, a place where scholars set their 
own agendas, von Neumann became one of the 
"scientifically anointed" at the Institute for 
Advanced Study at Princeton. 

(U) The first rumblings of war led the 
Institute and von Neumann to move far beyond 
their abstract academic origins. During World 
War II, von Neumann made important contribu- 
tions to the atomic bomb project. As a result of 
that involvement, he became entangled in the 
ENIAC computer effort at the University of 

(U) The University of Pennsylvania's World 
War II contract with Army Ordnance for the 
ENIAC had come almost by chance, just as the 
NDRC ended its computer program, and as firms 
such as RCA rejected pleas to turn their hard- 
pressed engineers to computer projects. 
Ordnance was in need of a way to speed the cal- 
culation of firing tables. With no other alterna- 
tive, the army accepted the proposal of two young 
engineers at the Moore School. They promised to 
build an electronic version of Bush's great 
Differential Analyser. Fortunately for the history 
of computers, John Mauchly and Presper Eckert 
were given a great deal of freedom and time. 
Their much delayed postwar delivery of the rela- 
tively special-purpose ENIAC was not treated as a 
sign of failure, and their plan for a programmable 

Page 236 



universal electronic computer was quickly 

(U) With the help of John von Neumann, they 
started the project (EDVAC) and began seminars 
that attracted the pre- and postwar generations of 
computer builders. 10 von Neumann's stature in 
the scientific and military communities had 
grown so much that his presence gave the Moore 
School's computer efforts the highest credibility. 
While working on the design of what is regarded 
as the first true universal computer, the EDVAC, 
the original leaders of the ENIAC project, 
Mauchly and Eckert, had become estranged from 
the university's administration and, to some 
degree, from John von Neumann. 

(U) Von Neumann, whose importance 
increased in the postwar years, also became alien- 
ated from the University of Pennsylvania. He 
decided to found his own computer initiative. He 
was soon able to convince his old academic home, 
the Institute for Advanced Study (IAS) at 
Princeton, to accept several military and civilian 
grants and to create a center to house his attempt 
to design and build his own computer. His "IAS" 
machine was intended to serve the needs of 
applied mathematicians and physicists. 

(U) Von Neumann did not confine himself to 
computer building. He became a major figure in 
Cold War science and policy. He advised all of the 
American leaders of the era, and he served on the 
most important science-related boards. He even 
became a good friend of OP-20-G and later NSA, 
serving on their expert panels. He gave them 
much technical and political advice throughout 
the 1940s and 1950s. His contributions included 
more than hints about new computer technolo- 
gies. He frequently urged the SIGINT agencies to 
sponsor fundamental electronic research to be 
conducted by leading academics." 

(U) While von Neumann was forging his Cold 
War reputation, the Moore School had begun its 
own machine, the EDVAC. Sponsored by Army 

Ordnance, EDVAC was to have the simplest of 
architecture. Although it was intended to be an 
operational machine for the Aberdeen Proving 
Grounds, it was also something of a testbed. A 
central goal of the project was to prove that a uni- 
versal machine could be made to work and to do 
it quickly. Therefore, EDVAC was designed as 
simply as possible. 

(U) EDVAC was a binary machine that 
depended upon a serial acoustic delay-line mem- 
ory. That memory was to hold both programs and 
data. The acoustic technology limited the 
machine to about 1,000 words of fast memory. 
Technological limits also dictated much of the 
EDVAC's internal organization. Trying to avoid 
the problems caused by the high failure rate of 
vacuum tubes, EDVAC's internal structure was 
made as sparse as possible. It had just one-third 
the number of tubes used in the ENIAC. 

(U) To keep the number of components at an 
absolute minimum, the machine had only a few 
built-in instructions. That was a wise decision. 
Each "instruction" demanded dozens of tubes 
and hundreds ofhandwired connections. And 
each increased the computer's cost and multi- 
plied the probability that it would experience a 
failure well before any significant computational 
task could be completed. 

(U) In addition to keeping the number of 
components to a minimum, EDVAC's designers 
limited the machine to the serial transmission 
and processing of data (one bit at a time). Serial 
processing also reduced the amount of failure- 
prone electronics. But it carried the price of slow- 
er processing rates. 

(U) EDVAC's designers made another trade- 
off that favored simplicity over speed. The 
machine's operations were based on "fixed clock" 
timing. That meant that no matter how little time 
one operation took, succeeding work had to wait 
until the next clock pulse. 


Page 237 


(U) EDVAC's planners tried to keep their task 
manageable by concentrating on building a 
machine for mathematicians. EDVAC was not 
intended to be a data processor. The EDVAC 
engineers did not try to solve the many problems 
involved in making input and output rates 
approach electronic speeds. Slow tape and card 
readers gave the machine its data, and its even 
more primal cardpunches and teletypewriters 
displayed results. Although one of the first com- 
puter programs written by the ENIAC-EDVAC 
group was for sorting, EDVAC's builders never 
pretended that it could replace tabulator equip- 

(U) While the EDVAC's designs were being 
set, the ENIAC's parents, Eckert and Mauchly, 
left the University of Pennsylvania and attempted 
to found and keep afloat their own for-profit com- 
puter company. After more than six years of anx- 
iety and tragedy, they completed the UNIVAC 

(U) The UNIVAC was also a delay-line, fixed- 
clock machine, but it went far beyond the EDVAC 
in terms of power and sophistication. One reason 
for that was the UNIVAC's attempt to become 
"the" new business machine, one to replace hun- 
dreds of tabulators. That called fcr the develop- 
ment of much-enhanced I/O technology. Anew 
data processing capability was to some extent 
achieved through the creation of magnetic tape 
systems, a development that helps explain why 
the first UNIVAC did not appear until 1951. 

(U) The goal of building a computer to replace 
the tabulators led to a very historic decision by 
Eckert and Mauchly. Because they wanted to 
maximize the speed of data processing, which 
typically demanded little calculation on a great 
deal of information, they deviated from a purely 
binary representation of numbers within the 
UNIVAC. It had what was called at the time a 
"decimal" representation. Although UNIVAC 
used binary circuits, a decimal format was 
imposed to speed the input-output functions. 

(U) Eckert and Mauchly's commercial com- 
puter aspirations, as well as John von Neumann's 
academic ones, were just emerging when the 
University of Pennsylvania decided to host its his- 
toric summer 1946 Moore School conference. All 
those who had made contributions to computing 
during the war were invited to hear presentations 
by von Neumann and others who were outlining 
the computers of the future. 

(U) A Summer in Philadelphia - an Exciting 

(U) It was probably Howard Campaigned 
attendance at an earlier (May) navy symposium 
that made him aware of the Moore School confer- 
ence. The Washington meeting was where he first 
made contact with von Neumann and where he 
realized that the general-purpose computer was 
going to be built, with or without OP-20-G. 
Campaigne decided that "G" should at least have 
a chance to be one of its sponsors. 

(U) He hurriedly arranged for some funds and 
then called his assistant, James Pendergrass, ask- 
ing him to attend the coming Philadelphia sym- 
posium. Campaigne was unable to tell him much 
about what was to take place in Philadelphia. As a 
result, much of what Pendergrass encountered 
surprised as well as thrilled him. 

(U) During the Moore School's summer pro- 
gram, Pendergrass studied the designs ofthe 
ENIAC and those for the much more advanced 
EDVAC. He listened to the presentations ofthe 
other men who had begun to develop universal 
electronic machines. 

(U) Coming from a physical sciences back- 
ground and being an advocate for applied mathe- 
matics, Pendergrass was especially taken with 
John von Neumann's ideas, including his version 
of a programming language. When von Neumann 
outlined the concept for his new Institute for 
Advanced Study machine, Pendergrass became 
convinced that "G" had to have a von Neumann 

Page 238 





type of computer. He thought the von Neumann 
design was the best, the one that would be the 
first to appear in hardware, and the one most 
likely to be cloned by a manufacturer. Well before 
the Philadelphia conference was concluded, 
Pendergrass convinced Howard Campaigne that 
OP-20-G should have one of the "new" comput- 
ers, specifically one with a von Neumann archi- 

(U) Pendergrass was not the only one who 
believed in the IAS design. The mathematical and 
applied physics community took it as the model 
for computers for the 1940s and early 1950s. The 
emerging von Neumann architecture was espe- 
cially attractive to mathematicians because it 
promised to provide a much faster and more pre- 
cise computer than other designs of the era. It 
had the potential to be faster than the serial type 
of machines by factors in the hundreds. At least 
five IAS computers were copied in American uni- 
versities and advanced research centers. 

(U) The IAS machine was not going to send or 
process data bit by bit, nor was it going to make 
one operation wait for a "clock." It was going to 
send data simultaneously and would initiate an 

operation as soon as the previous one finished. As 
or more important, it was not going to be based 
on the limited delay-line memory, von Neumann 
bet that RCA would keep to its pledge and devel- 
op the very advanced Selectron tube within a few 
months. The Selectron promised to be a fast 
memoiy that would maximize the potentials of 
electronic speed and parallel data-transmission 
and processing. 

(U) To many in the computer field, however, 
von Neumann seemed too far ahead of the tech- 
nology. He also appeared to be naive about how 
much of the complexity of his proposed machine 
could be mastered by his handful of engineers. 
His critics thought the EDVAC group was taking 
a more sensible course: creating a minimal and 
reliable computer that had a possibility of being 
completed on time. 

(U) Given the ambitious nature of 
von Neumann's computer, Pendergrass and 
Campaigne sensed that it was not going to be easy 
to persuade "G" that scarce resources should be 
devoted to a machine that was not yet fully 
designed, let alone built. No one, in fact, could 


Page 239 


predict when any of the new computers would be 

(U) Despite von Neumann's reputation and 
the accolades that the atomic energy community 
was awarding to the IAS design, Pendergrass and 
Campaigne knew they would have to prove that 
"a" machine could compete with al! the special- 
purpose devices that were in place at "G" as well 
as those that were being planned. And they would 
have to show, without insulting anyone, that the 
new computer would be as good as or better than 
Goldberg, the perhaps-universal comparator. 
Pendergrass and Campaigne were in a situation 
quite analogous to Hooper's in the early 1930s: 
How could they convince "operational" types that 
scientists had a better and practical grasp of the 

PES} Pendergrass got to work during the sum- 
mer and continued on, with Howard Campaigne's 
enthusiastic help, through the remainder of the 
year. They composed two persuasive technical 
reports. 12 

(TS} The first was sent to "G's" higher-ups in 
October, the second in December. A great deal of 
effort had gone into both reports to ensure they 
would convince the cryptanalysts that, like the 
proposed Sled, the von Neumann machine would 
end the horror of having to wait two years while a 
requested special machine was constructed. 13 The 
reports did not refer to any particular experience, 
but Pendergrass knew that many in his audience 
had gone through the frustrations of World War 
II when almost all the RAMs (and the Bombes) 
had arrived too late. 

f¥S} In one of the classic statements in the 
history of computers, Pendergrass wrote: "It is 
not meant that a computer would replace all the 
machines in Building #4, nor is it meant that it 
could perform all the problems as fast as the 
existing special purpose machines. It is however, 
the author's contention that a computer could do 
everything that any analytic machine in Building 

#4 can, and do a good percentage of these prob- 
lems more rapidly." 14 

(TS) The text of the report reflected both 
Pendergrass' orientation and the nature of mid- 
19408 computers. After explaining the logic of the 
von Neumann machine and admitting that it 
might be some time before any such computer 
would be available, he outlined what he thought 
was to be a standard programming language, one 
based on von Neumann's "one address" con- 
cept. 13 

(U) Von Neumann envisioned a machine that 
would be used for very precise calculation and lit- 
tle data processing. To speed calculating, it was to 
be a pure binary machine. To further improve 
performance, he had turned away from the origi- 
nal EDVAC idea of a four-address instruction. He 
had come to believe that the most efficient 
instruction format should include only one place 
to get or put data. That would allow, given the 
word size of the computer, more precise calcula- 
tions without additional hardware. 

(U) He argued that because his machine 
would be busy with much internal work, such as 
multiplication and division, it would be more effi- 
cient to have "get" and "place" addresses in sepa- 
rate statements. Only if a computer was to be 
used for much I/O and little calculation would a 
multi-address instruction be reasonable. 

(U) Von Neumann also believed that his 
machine should have very few commands. The 
smaller the number of commands, the less inter- 
nal circuitry that would be needed. Following his 
mandates, his engineers were able to reduce the 
number of components in the IAS machine. It 
had only two-thirds the number of tubes of the 
EDVAC. Von Neumann's mathematical focus also 
meant that he gave little thought to the I/O 
problem. What happened within the computer 
was more important to him than handling mass- 
es of data. 

Page 240 



(T9//gf) Pendergrass agreed with all of von 
Neumann's ideas, but in his reports he suggested 
that a few additions be made to von Neumann's 
set of minimal instructions. They were to be ones 
which, like multiplication and mod 2 commands, 
would be needed to meet special cryptanalytic 
needs. Especially important to him were those 
which would speed the analysis of baudot traffic. 

(=fi$) The politically important parts of the 
Pendergrass reports were the sections in which he 
and Campaigne presented computer programs 
for major cryptanalytic attacks. Demonstrating 
how the machine would perform the attacks was 
critical if'G" was to be persuaded to invest in a 
computer. Of course, the reports reflected an 
implicit faith that programming would be much, 
much less of a problem than building a special- 
purpose machine. No mention was made of how 
long it took to write the programs. 

(TS//SI ) The first report included software 
programs for a Generalized Copperhead problem, 
a Four-Wheel Enigma Grenade Problem, and 
a Hagelin attack. The December report was 
intended to impress any holdouts. It contained 
programs to imitate two of the grand achieve- 
ments of World War I, the Duenna and the 
Mercury. Asa concluding argument, the report 
showed how to end the great crypto-disappoint- 
ments of the war years. There was a program that 
could imitate a cipher wheel. It showed that a 
general-purpose computer might act as an elec- 
tronic Super Bombe. 16 

(U) Buy a Computer, Now 

(£} The reports made their point, at least with 
farsighted men like Joseph Wenger. He took 
action even before the Pendergrass-Campaigne 
report of December was completed. To reinforce 
Pendergrass' arguments, he immediately 
arranged for ERA's John Howard to conduct a 
computer feasibility study and assigned 
Pendergrass to continue to survey the computer 

(U) Pendergrass did not waste time. In 
November he informed "G" that many military 
and some civilian agencies were very interested in 
digital computers and that projects at the 
National Bureau of Standards and RCA (with the 
IAS) stood a good chance of producing machines 
by 1948. The IAS-RCA project, he thought, had 
the best design and best chance of success. 17 

(U) After attending another major computer 
conclave at Harvard in January, Pendergrass for- 
warded a new survey of America's and Britain's 
computer ambitions. He cited the emergence of 
more computer projects, most of which were sup- 
ported by government agencies. The navy's 
Bureau of Ordnance and the ONR, he showed, 
had already established quite a foothold, as had 
Army Ordnance. Even the Census Bureau had 
become involved. 18 Among the six active proj- 
ects 19 (the one under John V. Atanasoff had just 
been cancelled by the Naval Ordnance 
Laboratoiy), the IAS computer, Pendergrass 
reported, continued to be the best option. It 
remained much closer to completion than the 
proposed Whirlwind at MIT, and it was more 
suited to cryptanalytic work than the upcoming 

(U) The only nonpositive things that 
Pendergrass had to say about the Princeton 
efforts were that he had discovered that RCA and 
Princeton did not have a formal agreement bind- 
ing the corporation to build a computer and that 
its valuable Selectron was still in the "uncertain" 
category. Neither seemed critical to Pendergrass, 
however. He expected a working IAS machine by 
mid- 1948. He assured "G" that if the Selectron 
were not perfected, an electrostatic memory, such 
as the one proposed byMFPs Jay Forrester, 
would serve as a fully acceptable substitute. 

fS-) Pendergrass' surveys were read by 
Howard Campaigne, then sent to Joseph Wenger. 
Wenger trusted Pendergrass, and he believed that 
OP-20-G should gain a foothold in computers 
before one of the other branches of the navy 


Page 241 


established a monopoly. Without waiting for 
John Howard and C. B, Tompkins to submit their 
ERA report (it arrived in February 1947), Wenger 
made a commitment to acquire a von Neumann 
type of computer. *° 

( Tg//8f) Even before a contract was let, plans 
were laid to use the new machine on major oper- 
ational problems. 21 "G" even put aside the idea of 
building an electronic Super Bombe, at least until 
the potentials of the new universal computer were 

(U) In January Wenger was so enthusiastic 
that he ordered his men to establish project 
"Atlas," although he did not yet have the funds to 
design and acquire a machine. az The name 
"Atlas" was picked because a comic strip used it 
as a name for a "mental giant," but a reference to 
raw courage would have been as appropriate. 
Wenger still had to gain formal approval for the 
"G" computer. 

(•S^As Wenger struggled to find the money he 
needed, additional crypto-studies reinforced the 
initial enthusiasm, and went beyond it, perhaps 
raising expectation a bit too high: This opens 
tremendous possibilities in the field of clinical 
attack by speeding this attack up to the point 
where large volumes of traffic may be so 
processed. With sufficient skill in preparing the 
logical control, it seems possible that the machine 
may be made to perform any cryptanalytic opera- 
tion now done by hand, which does not require 
intuition." 23 

(■S} Wenger did everything he could to make 
sure the "G" computer proposal would be funded. 
He had Pendergrass assigned as a !iais3n to the 
Office of Naval Research. It was exploring com- 
puters and was intensely committed to furthering 
applied mathematics. With Pendergrass in touch 
with ONR's experts, they would be unlikely to 
block the "G" request on technical grounds. Other 
mathematicians in the agency were sent to 
important computer seminars: Eachus, 

Campaigne, Blois, Tordella, and others met with 
the "greats" of computer history, such as Alan 
Turing and M. V. Wilkes. 

dSfr The contacts and investigations soon 
started to pay off. "G" was gaining a reputation as 
one of "the" centers of computer expertise. Other 
development projects, such as Whirlwind at MIT, 
gladly shared design information. 

(U) At least in terms of computer architecture, 
"G" was well integrated with high science. "G" 
became committed to the atomic scientists' 
favored way of sending data within the machine: 
all the bits at one time in parallel, rather than one 
bit at a time (serial mode) as in the EDVAC. As 
important, "G" wanted Atlas to have a single 
memory, one to hold both data and instructions. 
That was in contrast to some architecture, such as 
those of Howard Aiken at Harvard, who thought 
separate memories, concurrent processing, and 
dozens of registers made for a more powerful 
computer. 24 Without any hesitation, "G" favored 
a pure binary system for its computer. The idea 
that became embodied in the UNIVAC, that some 
decimal representation was more efficient was 

(U) While his research crew defined Atlas, 
Wenger worked on the politics of acquisition. He 
convinced the CNO of the need for Atlas, gained 
an extremely high priority rating for it, 25 and then 
sidestepped some serious objections from the 
Bureau of Ships. 

(U) In response to hints there were already 
enough navy computer efforts and that long term 
research should be left toothers, Wenger 
informed the bureau that "G" needed to acquire a 
"special analytical machine." The word "special" 
gave OP-20-G the opening it needed to avoid a 
worst-case situation in which it would be forced 
to wait for and accept a machine it might not 
want. It also gave "G" the chance to play a positive 
role in the emergence of the computer industry. 26 

Page 242 



EO 3.3(h)(2) 
P.L. 86-36 

(U) Well before authorization had been grant- 
ed, "G" began a more detailed design and made 
evaluations of possible computer manufacturers. 
With all the other government agencies sponsor- 
ing research in the field and with the interest 
shown by several private companies, "G's" 
experts did not anticipate that a large investment 
would be required for the design or for the hard- 
ware. "G" still thought RCA would enter the mar- 
ket. The National Bureau of Standards also 
seemed ready to build a computer. Wenger 
expected to have Atlas at the Nebraska Avenue 
complex in approximately two years. 27 

(U) Whatever the options, Wenger wanted 
quick action. Even though a "special" machine 
had been approved and although Monogram 
funds were available, there was always the chance 
that the White House might decide that comput- 
ers were a luxury. Even the $100,000 to 
$300,000 for the machine might be seen as too 
much for a peacetime intelligence agency. 28 

(U) Laying out the general specifications for 
Atlas was relatively painless. Pendergrass had 
done his technical homework, and his recom- 
mendations were only refined, not changed. 
Beginning in March 1947, when "G" decided to 
take more responsibility for designing its Atlas, 
Campaigne, Eachus, Pendergrass and many oth- 
ers at "G" began to meet to detail the functional 
characteristics of their newest "analytical 
machine." They even began to write programs. 
The enthusiasm was so great that many worked 
nights and weekends on their problems. 29 

■{S} "G" was to have a von Neumann comput- 
er, not asouped-up version of its older devices. 
Suggestions that Pendergrass' original sketches 
be altered by adding special-purpose attachments 
were adamantly rejected, as were those recom- 
mendations that the machine have control 
switches and plugboards. Software, driving ele- 
mental circuits, was to be the only control mech- 

{S}- But the number of commands built into 
the machine was to be expanded beyond von 
Neumann's original list, and Pendergrass' early 
recommendations. By 1947 close to forty com- 
mands were in the design. The expansion was 
aimed at easing cryptanalytic processing, as had 
been the alteration in the fundamental word size 
in the machine to six digits. That would allow let- 
ters as well as numbers to be analyzed. 

(9} The additional commands were at the fun- 
damental level of the machine. There were \no 
suggestions that complex sequences to imitate 
entire processes be wired into Atlas. A series of 
multiplication commands and a divide instruc- 
tion were included, however, as were shift com- 
mands and noncarry arithmetic capabilities* 
Shifts were especially us eful when rotor or wheel 
stepping was required. 



Also, there were hopes that a random number 
generator could be devised. 

(U) Campaigne, Eachus, and the others on the 
design team had bright hopes for Atlas. But there 
were limits to the aspirations. They accepted the 
fate that plagued the first computer generation: 
They did not attempt to write a compiler or a 
high-level language for the machine. The only 
treat the "G" group gave programmers was the 
luxury of writing in octal rather than binary nota- 
tion. That provided some relief, but it did not 
allow a programmer to avoid specifying the loca- 
tion of memory addresses in "absolute" terms. 
There was no software to automatically keep 
track of where instructions or variables were 

(U) Like the TAS computer, Atlas was to be 
centered about what the Princeton group consid- 
ered "the" solution to the memory problem, the 
RCA Selectron tube. It would allow an electronic- 
speed mass memory, something needed to meet 
the potential of parallel data transmission and 
processing. Hopefully, the Selectrons would sup- 
port a large memory. In 1990s terms, Atlas was to 
have 64K. In terms of the longer word size of the 


Page 243 


Atlas, that was equal to 16,384 "cells." That was 
orders greater than what was planned for 
EDVAC. 3 ' 

(U) The Selectron was under development at 
RCA's research laboratory. Rachman's tube 
promised to be much more powerful than the 
other types of binary electrostatic storage devices 
that were under development. And it was expect- 
ed momentarily. Although some at OP-20-G had 
treated many of Rachman's ideas as more than 
fanciful, because of his advanced work during the 
war, he had become an ally of von Neumann, and 
his work demanded respect. 32 

(U) The Selectron was a complex device, but it 
had a great advantage; it was small and fast. Its 
size was one of its great attractions because other 
high-speed memories of the period, such as delay 
lines or the Wilkes electrostatic tube, 33 took a 
great amount of space. Unfortunately, the 
Selectron proved to be too complex. 

(U) It was based upon the principle that "an 
insulated secondary-electron emitter can be 
made to 'float' at either of two stable positions..." 
Deceptively simple, the principle demanded 
much delicate hardware. Inside the three-by- 
seven-inch tubes was a dielectric target that was 
divided up by sixty-four metal bars and sixty-five 
circular metal rings. They created 4,096 "cells" 
that were the storage areas. When the four walls 
of a cell were all more positive than some partic- 
ular voltage, a "bit" was registered. 

(U) To von Neumann's and "G's" great disap- 
pointment, all that was too much, even for the 
great Jan Rachman. 34 By spring 1947 RCA had to 
admit that it might be some time before the 
Selectron was ready. That led to some technolog- 
ical soul searching in Princeton and Washington. 
The IAS put more effort into a television-like elec- 
trostatic memory and even explored the possibil- 
ity of ultra-high-speed secondary memory based 
on magnetic wire wound on bicycle wheel drives. 

(U) The news about the Selectron was only 
one indication that the computer revolution was 
going to take much longer than had been thought. 
RCA began to make it clear that it was pulling 
back from its hints of becoming a manufacturer, 
the National Bureau of Standards program had 
slowed to a crawl, and the probability that the ex- 
ENIAC team, Eckert and Mauchly, could deliver 
their promised computer to the Census Bureau in 
time for the 1950 census sank to near zero. 

(U) By spring 1947 Atlas was on its own. If "G" 
were to have its computer, it would have to take 
even more responsibility, perhaps even for a very 
expensive failure. And it would have to make a 
critical technological choice. 

(U) Little Thanks for Tliat Memory 

(U) In April 1947, after learning about the 
faltering industrial commitments and the 
Selectron's possible stillbirth, "G" made two very 
significant decisions. The first was to continue 
with the project and the acquisition of a comput- 
er despite the absence of an "industry" or even a 
university that seemed willing to build comput- 
ers. The second decision was perhaps more dra- 
matic. 35 

(T0//3I) When it was learned that the 
Selectron would not be available, there was a crit- 
ical meeting at "G's" Nebraska Avenue headquar- 
ters. Some of those in attendance thought that 
without the high-speed memory it would be 
senseless to continue more than very general 
design work. What use would Atlas' electronic cir- 
cuits be if the memory was a slow tape or similar 
device? Even looking for a manufacturer for Atlas 
did not make sense to them. There were a few 
suggestions that the entire project be put on hold. 

(TC//9F ) Howard Campaigne, perhaps wor- 
ried that such a decision would end chances of 
funding, put up a stubborn fight. He won half his 
battle: The work was not canceled. But his victo- 
ry seemed to open the door to some dangerous 

Page 244 



possibilities. His recommendation to go with 
what had always been the "fall back" memory for 
Atlas and Goldberg, 36 a magnetic drum, stood the 
chance of making Atlas and "G" look rather fool- 
ish. It could make Atlas very slow and perhaps 
very dumb. 

£ Rj//flI) Drums were much faster than tapes 
or cards, but they delivered information at a rate 
of i/400th or less of delay lines. Some estimates 
of the period gave the Selectron and electrostatic 
memories a 1,000-fold advantage. 37 If microfilm 
could have been made to be "rewritable," it could 
also have made a drum look antiquated. Seventy- 
millimeter microfilm held 12,000 bits per inch; 
drums had a density of from 100 to 200 bits. 38 

GBS-) Although "G's" RAM group realized that 
such a memory would slow the proposed machine 
manyfold, by a close vote its members decided 
that a drum would be acceptable. It seemed a 
much better choice than postponing the project 
and being left dependent on the whims of an 
almost nonexistent computer industry. 

-£BS} Campaigne and his associates realized 
they were taking a chance. There were hosts of 
mechanical as well as magnetic-electronic chal- 
lenges to overcome. Whether the "drums" were 
long bars or three-foot "wheels" covered with 
magnetic tape or sprayed with a magnetic coat- 
ing, the problems of milling, sensing heads, and 
drive motors remained unsolved. Even ERA with 
a head start on drum construction because of its 
connection to the earlier RAM projects, did not 
have a finished and sure technology in hand. ^ 

(=F8> "G" decided to take the risk. While the 
IAS group waited for the Selectron's development 
or the appearance of another electrostatic memo- 
ry, "G" started to work on revised designs for a 
drum machine. It also began a search for some- 
one to build the newly defined Atlas. 40 

CfS) No serious consideration seems to have 
been given to having, as would many atomic ener- 

gy research groups, a university take charge of 
final design and manufacture. And "G" did not 
spend much time investigating the few compa- 
nies that seemed willing to build computers. 
Thus, soon after the critical April 1947 meeting, 
ERA was chosen even though "G" knew how busy 
the young firm was with its first contracts. 

(8) There was some worry that Atlas might be 
a bit too much for the new company and that 
some emerging problems with magnetic drums 
might not be conquered. 41 But in August 1947 
ERA was given a design contract. And it was 
informed that "G" wanted a machine soon. ERA 
was not to wait for the results of the several 
research projects OP-20-G and the SIS were 
sponsoring to develop multifunction and ultra- 
high-speed tubes and new circuits. And there was 
no thought of delaying Atlas just because there 
were not yet any high-speed printers suitable for 
an electronic computer. 42 

(325) There were a growing number of reasons 
why "G" wanted ERA to quickly prove the worth 
of a universal machine for cryptanalysis. Just as 
ERA was put to work on the final designs, the 
Sled project with its special architecture was 
being launched with much support from the 
Bureau of Ships. In addition to having some com- 
petition, Atlas had to face another possible trau- 
ma; there were well-grounded rumors that the 
Monogram budget was to be cut severely so. 41 

fFS} With a great deal of help from "G's" 
research group in Washington, ERA was able 
to develop an acceptable design within a 
few months. As requested, it matched the 
von Neumann concepts and was aimed at avoid- 
ing manufacturing problems. Some rather useful 
ideas were sacrificed to the needs of the produc- 
tion schedule. A second processor, which would 
check results, was not included, and the sugges- 
tion to develop a partitioned memoiy was 
dropped. Having as many as eight active "accu- 
mulators" was also regarded as too much of a 


Page 245 

T O P W Pl Hfninimmi-i nr. ..,,» m{ ^ flMp M7U/ri 

fES} In early spring 1948, in return for prom- 
ises to use as much standard equipment as possi- 
ble, ERA was awarded a construction contract. 
There was a caveat, however. ERA was more than 
encouraged to build Adas in a way that would 
allow the substitution of electrostatic (Selectron) 
storage if and when it became available. 

(T0//0I) ERA and "G° were in a hurry. Adas 
was given an AA priority, ERA borrowed much 
from MITs Whirlwind project, and ERA gave 
Adas as much attention as possible even when it 
had to rush to complete some special-purpose 
machines to attack Russian targets. 44 

(S> While the engineers in St. Paul were work- 
ing on Atlas, the mathematicians-turned-pro- 
grammers in Washington built their own com- 
puter to prepare for Adas' arrival. They wanted 
programs ready to help prove their electronic 
computer's operational value as soon as it was 
delivered. Constructed within four months out of 
relays and a small magnetic drum developed by 
ERA their Abel computer was a logical clone of 

Adas. It gave "G" more than a year's head start in 
training programmers and in writing some oper- 
ational programs. Its drum was not large enough 
to perform all of Adas' chores, and its relays were 
hundreds of times slower than ERA's circuits, but 
it came to be almost a "pet" of the research 
group " 


P Atlas 1 

(•G-) Meanwhile, despite the growing pressures 
on ERA, it was able to work something of a com- 
puter miracle: Adas was delivered to the navy in 
early December 1950, fairly close to the anticipat- 
ed delivery date. It had taken ERA less than two 
years to construct the machine, perhaps because 
so much time had been spent preparing for its 
production stage and because of ERA's experi- 
ence building special-purpose machines, such as 
Goldberg. In fact, Adas was the thirteenth project 
for "G." 46 

(€T Most of the design goals were accom- 
plished. That made Atlas one of the very first 
operational computers in the world. ERA also 
achieved another sort of computer first: Adas 
worked and worked well for 
; a decade after it was sent to 
Washington. A very effi- 
cient testing and mainte- 
nance schedule allowed 
replacement of tubes before 
they caused an unexpected 
failure. That contributed to 
an almost unheard of 90 
percent "up-time" (avail- 
ability), which made ERA 
very proud and very anx- 
ious to transfer its new 
computer skills to the com- 
mercial marketplace. 47 It 
was also proud that it could 
have built one of the most 
powerful of all the early 
computers using only 2,700 
tubes and that its drum per- 
formed reliably- 48 Asa 
result, ERA and its follow- 

Page 246 


10^ afcUKb l /ZlUM I NlimfcL U&A, AU3, CAM ODR AMD HZU/X1 

lEO 3.3(h)(2) 
P.L. 86-36 

on companies became leaders in magnetic drum 
technology and gained a reputation as supercom- 
puter builders. 

(U) Saving a Reputation through Logic 

PR8} But all the original Atlas goals were not 
attained. The machine had cost perhaps three 
times the early postwar estimate; its delivery 
price was just short of $1,000,000. 49 More 
importantly, the drum held less than one-third 
the amount of information that had been hoped 
for in 1947. But ERA was able to rotate it at an 
extremely high speed. Partly by reducing its size 
from the dimensions of earlier drums (three feet 
in diameter) to twenty-five inches long and eight 
inches in diameter, Atlas' drum was ten times as 
fast as the one installed on Goldberg. "" The 
increased speed helped, but it did not solve the 
memory access problem. The 1950 Adas began its 
life as a very slow machine because the program, 
as well as data, had to be read from the drum. 
There did not seem to be a viable technical save. 
Replacing the drum with the still expensive and 
irritable electrostatic or delay-line memories 
seemed impractical. 

-ffS) The programmers atOP-20-G were 
charged with finding the best solution they could. 
Perhaps to everyone's surprise, they came up with 
an answer that made Adas competitive with other 
computers of the time. 

i^T The solution they devised was called 
"interlacing." Combined with very careful pro- 
gramming, it increased Adas' speed by a factor of 
more than 300. That meant that the drum-based 
Adas became approximately two-thirds as fast as 
a similar machine using the new magnetic core 
memory of the mid-1950s. In fact, Adas came 
close to being a match for the IAS machine. 5 * The 
increase in Atlas' speed came at a high cost to the 
early programmers, however. 

(S^The trick they had to pull off was to place 
instructions around the drum in such an order 

that rotation time until the next expected instruc- 
tion was minimized. At first, \ a plugboard was 
used to accomplish the necessary scrambling of 
once-sequential locations. By 1951 an automatic 
"dial" system was installed that eliminated the 
need to replug a board for each program. 
Although the relocation of instructions was made 
automatic, programming was; not. To utilize the 
"interlaced" instruction, programs had to be 
drafted on large sheets ofp4per. The two-by- 
three-foot sheets allowed the programmers to 
keep track of where the instructions were located 
and allowed them to perform jtiming miracles so 
that the call for the instruction came when the 
drum was in the correct positions. 52 

(T8^-Despite the near agony of Adas program- 
ming, a wide range of statistical attacks was run 
on the machine. It proved Pendergrass' point 
about flexibility although writing Adas' pro- 
grams, took much, much longer than had been 
imagined by him in 1946. Such fondly remem- 
bered programs as Bootstraps for the identifica- 
tion of no nrandom distribu tiqhs (roughness test- 
ing) and I *~ 

analysis aids took 

enormous human effort. 53 

(U) The new programmers in Washington 
could not find any way to compensate for anoth- 
er of Atlas' failings, however. Adas could not be 
coaxed into becoming a data processing comput- 
er. Its input-output capabilities were too limited. 
It brought data to its drum through a photoelec- 
tric papertape reader; an Electromatic typewriter 
and a tape punch handled its output. T^eTwere 
no pathways for punch card machines nor for the 
just emerging magnetic tape drives. 

(*ft^ A severe limitation with many ramifica- 
tions was Adas' inability to put the input tape- 
reader under program control. All the reader 
could do was load the drum. The paper tapes 
could not be used as a dynamic source of data. 54 

fS4 Adas' sparse I/O was a result of conscious 
design judgments, ones which mixed technologi- 

top oconrn/ooMiNT/mBL uoa, auo» ca n obr a m d n zu/xi 

Page 247 


cal possibilities with operational needs and with a 
strong dose of "GM's" goal of fulfilling Wenger's 
dream of "mathematical" cryptanalysis. 

(S> When Atlas was being designed, the only 
new large-scale secondary memory media that 
seemed to have input potential was magnetic 
tape. But tape systems with the possibility of 
holding massive amounts of information 
remained in the development stage and were 
proving very stubborn. To wait for their matura- 
tion would have delayed Atlas' construction; to 
attempt to anticipate what circuitry Adas needed 
to hook up to future systems would have been 

4$) Attaching Atlas to an IBM card reader 
might have seemed attractive atone point, but 
there may have been questions about using that 
company's equipment in a competitor's system. 

tS^In any case, "G" had decided by 1948 that 
energies would goto increasing Atlas' internal 
processing power by adding additional "instruc- 
tion" circuits. The expansion of the number of 
hardwired instructions was intended to encour- 
age the use of K mathematical"cryptanalysis. The 
many binary multiplication and shifting instruc- 
tions and the divide circuitry made Adas more 
expensive and harder to manufacture, but they 
speeded statistical testing by many factors. 

(S)- Those features seemed so attractive that 
despite the severe I/O limitations, a second Adas 
was ordered some six months before the first was 
shipped from St. Paul. Because the initial model 
absorbed the development cost, this Adas was 
priced at one-third of the original. At the same 
time, a new design cycle was begun. The Adas II, 
which in its civilian guise was called the 1103, was 
the machine that anchored the ERA group's com- 
puter building reputation, although its original 
price was $1,250,000. w 

Q tS//0f ) For a time, however, it was thought 
that Adas III would be a data processor as well as 

a "number cruncher." But Adas II could not over- 
come the earlier I/O limitations. When it was 
being designed, the Raytheon Company 
announced it was perfecting tape drives. ERA's 
engineers built a program-controlled I/O feature 
into the machine only to discover serious techni- 
cal difficulties with the Raytheon magnetic tape 
systems. Despite a last-minute effort, eleven were 
delivered without a tape capability. 56 But its 
internal processing powers were much enhanced. 
Some electrostatic storage was added, the drum 
was improved, a two-address logic was intro- 
duced, the word size was increased, and several 
very useful basic instructions were added. 

(U) All in all, "G" thought that its sometimes 
frustrating computer adventure had been worth- 
while. The agency's advocates for general-pur- 
pose machines had made their point, and the 
work on Adas helped to establish ERA as a com- 
puter company. The Adas designs and designers 
would play an important part in the history of 
automation of communications intelligence. 

(U) The relatively happy ending of the Adas 
project was not quite matched in the army's 
attempts to establish its place in the computer 

(U) The Army's Problem 

4G) The SIS lost many of its engineers after 
the war, and it was unable to create its own ERA. 
In response, it planned to do some machine 
design in-house, go to contractors for details and 
components, and, when necessary, assemble its 
secret special-purpose devices itself. That seemed 
quite efficient, but the SIS was unable to follow 
that approach when it began its quest for its first 
general-purpose computer, a machine that took 
its name from another newspaper cartoon char- 
acter, "Abner." Although some have claimed the 
name was chosen because the modern computer 
without a program is a dumb machine, the selec- 
tion of the name in 1949 may have been inspired 
by the nature of the search for a design and a 

Page 248 



manufacturer. The army's codebreakers had gone 
through a bizarre and agonizing odyssey during 
the previous three years. The experience was 
almost cosmic. 57 

fj9) In 1945 there was a significant reduction 
in force at the SIS. But Solomon Kullback and Leo 
Rosen were able to retain enough personnel and 
funds to continue the old "F" branch - one with a 
fresh bureaucratic name and expanded powers. 58 
One of their first and most important decisions 
was to appoint one of their young proteges to 
head a new subsection. In January 1946 Samuel 
S. Snyder, who later became a major figure in the 
computer and information world, was asked to 
survey the wartime computer developments and 
then to turn his new RAM research group into a 
dynamic force. The group was to keep the agency 
informed about all the computer developments in 
the world and to act as an advocate for further 
automation within the agency. 

QB) Snyder and his coworkers, Mary Roseboro 
and William May, began by gathering and rewrit- 
ing all the documentation that remained on the 
World War II machines at SIS. Soon they decided 
to expand the survey to include the devices OP- 

20-G had obtained. That turned their project into 
creating what became the irreplaceable "Machine 
Aids to Cryptanalysis" series. 59 

■tfif Snyder did more than follow in-house 
developments. By mid- 1947 he had visited the 
National Bureau of Standards and had made 
some contacts with those in academia and the 
commercial sector that had interests in computer 
development. But he got somewhat of a late start 
at his attempts to bring a computer to the agency. 

(S//ST) The SIS was at least a year or two 
behind OP-20-G's computer work. One reason 
was that someone like Snyder had not been 
selected to attend the Philadelphia computer 
symposium in mid-1946 that had so impressed 
Pendergrass. Instead, Kullback had sent a SIS 
engineer who had little or no mathematical or 
cryptanalytic experience. The man was not 
impressed with what he heard and did not report 
back to the SIS that a great technological 
and cryptanalytic opportunity had appeared. 
According to Snyder, the man did not even sub- 
mit a report on the ENLAC and EDVAC designs. 60 

(S//SI) Fortunately for the agency, Sam 
Snyder encountered Pendergrass' report. 
Inspired, he began contacting others who were 
developing what later became known as "comput- 
er science." Snyder became somewhat of a com- 
puter "trekkie." He attended all the meetings, 
such as the famous one at the Aberdeen Proving 
Grounds, of the just forming Eastern Association 
for Computing Machinery. 61 He learned of the 
vast data processing center at the Prudential 
Insurance Company (some 700 IBM machines) 
and the plans of one of its aggressive young exec- 
utives to make the Prudential a center for elec- 
tronic computer development and applications. 
That contact with E. C. Berkeley reinforced 
Snyder's rather philosophical view of the coming 
computer revolution. Berkeley was one of the first 
"futurists" in the computer field. 62 

(U) Samuel S. Snyder 

I UH fc>blKb l »UJ I V I IN i ;m E L U&A, AUS, LAN QBE A N D NZU/X I 

Page 249 


TS/7^JThere was a much more practical side 
to Snyder's trips around the East Coast. By at 
least mid-1947 the SIS decided that it would not 
let OP-20-G be the only one with an electronic 
general-purpose computer. Snyder was sent to all 
the companies and academic institutions that had 
indicated they were going to take the risk of build- 
ing a computer. The only centers Snyder did not 
seem to visit were ERA, Harvard, and Western 
Electric. He skipped ERA because he already 
knew about its computer design and, perhaps, 
because from the beginning, the SIS did not want 
just a clone of the navy's computer. Nor did it 
want to become dependent on the navy's captive 
corporation." He bypassed Harvard, where 
Howard Aiken continued to build ordnance com- 
puters, because the unique Harvard architecture 
did not seem right for cryptanalytic work. The 
Western Electric postwar devices also seemed to 
be a bit "old" in terms of design and hardware. 

(0//0I) S nyder's trips were exciting; he was 
becoming a pioneer. He was able to see all that 
the fledgling UNIVAC group in Philadelphia was 
doing. He even spent time with Grace Hopper, 
who was becoming a legend in the computer 
world for her contributions to programming. 
While in Philadelphia, he also made contact with 
the EDVAC team. Then he headed for Princeton 
and the Institute for Advanced Study where 
Julian Bigelow was leading the group that was 
slowly making John von Neumann's concept turn 
into hardware. The Institute and Bigelow were 
impressive, but Snyder continued his search. 
Soon he was in Boston, where he found what he 
thought was the most promising of all the com- 
puter projects in the country. It was the one with- 
in Vannevar Bush's old company, Raytheon. 
Bush, however, had nothing to do with the proj- 
ect. In fact, many of its bright engineers had come 
out of the World War II computer projects run by 
Howard Aiken at Harvard. 

(U) Snyder thought that Raytheon's R. M. 
Bloch, R. V. D. Campbell, and M. Ellis were doing 
the most exciting work in the country and were 

the most likely to be able to construct the type of 
computer the SIS needed. 63 In addition to the 
design of its computer, Raytheon was attractive 
because it was the only large corporation in the 
country willing to subsidize computer develop- 

(U) The company was in a unique position. 
Raytheon had gone from a moderate size firm in 
the 1930s to become a major defense contractor 
during World War II. It was aggressively seeking 
new products and markets that would allow it to 
keep its position in the peacetime economy. That 
was in contrast, for example, to IBM, which was 
unwilling to endanger its major product lines by 
leaping into computers. 

(T8//8i ) The small firms that were showing 
interest in computers were not viable alternatives 
for Snyder. Investing in them posed a risk for any 
purchaser. Even the one with the best reputation 
was showing signs that it was overreaching itself. 
UNWAC's Eckert and Mauchly wanted the SIS's 
work, but could not commit to building a 
machine that would suit the needs of cryptanaly- 
sis within a reasonable time. In addition, there 
may already have been security problems at the 
company. 64 

(S3 That left Raytheon as almost the only 
option for the SIS. Then the proposed Raytheon 
machine received more acclaim. While Sam 
Snyder was making the rounds of the computer 
centers, his research group had been examining 
the designs of all the proposed computers and 
found that Raytheon's was to be a data processor, 
at least much more so than Atlas or any of the von 
Neumann machines, because it was a four- 
address device. In addition, although it was a seri- 
al processor, its speed would be more than ade- 
quate because it was to have a large and fast 
memory built of Selectrons. Although ERA had 
concluded that the Selectron might never appear 
and turned to the drums, the SIS bet that it would 
soon emerge from RCA's research laboratory. 

Page 250 



-fS) Talks were begun with Raytheon's man- 
agement, and by fall 1947 the SIS group thought 
an advanced computer was about to be built for 
them. The early plans included more than just a 
fast memory. Raytheon promised to make its 
computer more of a data processor than any other 
by developing revolutionary magnetic tape and 
wire systems. They were to provide high-speed 
bulk input, and there were even hopes of devising 
high-speed output mechanisms. Perhaps as 
important, the SIS thought that Raytheon might 
deliver a machine before Atlas could be sent to 

^ It wasn't too long before those hopes were 
dashed. Raytheon let the SIS know that because it 
had obtained a contract for a computer from 
another navy agency (through the NBS), it would 
beat least three years before a SIS computer 
could be completed. Although the company 
offered an attractive price, $350,000, and indi- 
cated it would be able to provide the SIS with a 
machine that included their very promising plas- 
tic tape systems, they declared they would pro- 
vide it on their schedule. 

-(SHThe SIS group hesitated before accepting 
the new offer, thinking that three years was much 
too long to wait. They had been very busy writing 
their own version of the Pendergrass report and 
had already written programs for the type of 
machine they desired. Those investments seemed 
too much to waste. But there was no alternative to 

£ES} When the SIS managers returned to the 
company, they were somewhat resigned to a long 
wait, although they planned to bargain over deliv- 
ery schedules. As they started the negotiations, 
they received a shock. The company had 
reworked its estimate of the cost of a computer 
that met Snyder's needs. Raytheon now wanted 
so much more that the SIS turned the offer down 
without further bargaining. 65 

(U) Stratton's Dream Revisited 

(U) In near desperation, Snyder went to the 
National Bureau of Standards. Standards had 
visions of becoming what Stratton had desired a 
generation before: the center for computer devel- 
opment in the nation. John Hamilton Curtiss, an 
applied mathematician with a Harvard degree 
and wartime navy experience, had been hired by 
the new crusading leader of the NBS, E. U. 
Condon, to accomplish that. By 1947 Condon had 
funds for computer development and was on the 
way to becoming an intermediary for all govern- 
ment agencies' computer purchases. With a 
group of energetic engineers and mathemati- 
cians, the NBS guided, for example, develop- 
ments atUNIVAC and Raytheon. It was doing 
much more. It was encouraging and coordinating 
the work of many contractors who were develop- 
ing computer components, and it was making 
suggestions to builders to improve computer 
architecture. 66 

(U) But contact with the NBS did not lead to 
any immediate relief for Snyder. The best the 
NBS said it could do was to allow Snyder's group 
to attend the computer lectures it was conducting 
and to provide leads to new companies that might 
be willing to build a computer. 

(U) It was mid-1948 and the SIS still did not 
have a final design or a contractor. 

(6} The SIS went in circles for a time, then 
came back to an earlier contact, the Reeves 
Instrument Corporation, a New York firm that 
was a leader in the analog computer business. It 
had just completed a very useful and pathbreak- 
ing electronic differential analyzer, the REAC. It 
had also gained some digital experience by help- 
ing the University of Pennsylvania with its com- 
puter projects. More importantly, it had let it be 
known that it was going into the digital computer 
business. To do so it had hired one of the most 
unusual men in the early computer business, 

top Gr.Gnr.T//oo MiN T//rcr.L uoa, auo, ca n odr a n d n zu/x i 

Page 251 


Samuel Lubkin, to supervise the design and con- 
struction of its proposed REVAC. 

c$) Lubkin was an alumnus of the important 
University of Pennsylvania projects, and he want- 
ed to build his own improved version of an 
EDVAC. His past experience and the preliminary 
design of the Reeves machine convinced the SIS's 
team to support Lubkin's design although he 
planned to use delay-line rather than faster 
Selectron memory. Abandoning the Selectron 
was difficult, but there was an attractive trade-off: 
Reeves was proposing to build its advanced ver- 
sion of the EDVAC within one year for a bargain 
price of $150,000. 68 Although the Reeves 
machine would be serial and clock-based, it was 
to use four addresses; most importantly, it would 
beat Arlington Hall before the end of the 
decade. 69 

GS) Serious talks were begun with Reeves 
in early summer 1948 and Snyder and his team 
felt vindicated. 

OES) Then chaos took hold again. Just as 
negotiations were begun, Reeves announced it 
was not going to branch out into digital comput- 


tS) Lubkin immediately left the company. At 
first Snyder thought that all was not lost. There 
seemed a chance that Lubkin could become an 
SIS employee. At least the design for his machine 
could be finished. Talks were held: then Lubkin 
decided he wanted to found his own computer 
company. For a moment it appeared that the SIS 
might have its own version of ERA That was a 
short-lived dream. Lubkin could not raise the 
necessary financing. Lubkin gave up and took a 
government job, but not with the SIS. He went 
with the National Bureau of Standards. 

(U) There was no one left to build "Abner." 
For the SIS there was no American computer 
industry. Then there were a few moments of relief 
when the NBS gave some indications that it might 

arrange for a computer for the agency. As hap- 
pened so many times before, the hopes were 
defeated. The contractors the NBS was depending 
upon for the computers to be used at its impor- 
tant applied mathematics centers could not meet 
their schedules. They got so far behind that the 
NBS decided to build a machine for itself. That 
eliminated any chance that the SIS could get a 
machine within the near future. The NBS would 
be too busy arranging for its "interim" computers. 
Its work force had to concentrate on a machine to 
save the NBS's numerical centers, and its needs 
would keep available subcontractors busy. 71 

{$) Swallowing a great deal of pride, the SIS 
made another brief attempt to get Raytheon to 
reconsider. No deal could be struck. Then the SIS 
decided that it had only one alternative, unless it 
was to give up on the idea of being one of the first 
members of the world's "computer club." No mat- 
ter what the risk, it had to build its own machine. 
Snyder reasoned that since the SIS now had an 
engineering staff of some sixty men who had 
already had much electronics experience (as a 
result of their work in radar and other military 
electronics), there was a chance of success. 

"TC-) But Snyder knew his group required help; 
it needed some experienced and skilled computer 
designers to flesh out the functional sketches 
being produced by the SIS engineers, such as Ray 
Bowman and Dwight Ashley. Again the SIS went 
to the National Bureau of Standards. Since 
Lubkin was there and since he had already put so 
much work into the design ofthe REVAC, it 
seemed reasonable to expect cooperation and a 
detail design within a short time. The NBS was 
reluctant to take on any more responsibilities, 
however. They had their own crisis to deal with. 
But after emphasizing that it was too busy to 
build a machine for the agency, in early fall 1948 
the NBS offered to take on the design task for 
$150,000. 72 That seemed a bit too expensive. 
That was the amount that Reeves was to have 
charged for a delivered machine, but the SIS's 
options were limited. So, even though the NBS 

Page 252 


EO 3.3(h)(2) 
£.!_. 86-36 


was indicating that it was going to provide the SIS 
only with something close to a copy of the very 
simplified delay-line SEAC, it was rushing to 
build for itself, the offer was accepted. 73 

4S1 To prepare for the arrival of the design, a 
team of engineers was formed and programming 
classes were begun. Snyder felt that his ordeal 
was finally over. The project seemed about ready 
to contribute operational results; there was great 
enthusiasm. The new programmers went beyond 
th eir lessons and began to write routines to attack 
Soon the mathematicians 


at the SIS were swept up in the excitement over 
the about-to-appear computer. Dick Liebler and 
Hugh Gingerich even d evised anew class of 


ones that could be done 

only on a high-speed digital machine. 74 

tS} Unfortunately, the plans did not come 
from the National Bureau of Standards on sched- 
ule. Its crew was so busy with the bureau's own 
computer problems and those of the contractors 
it was supervising for other government agencies, 
that all Snyder got from them were promises to 
hurry. The situation got worse when Lubkin 
decided that Standards was not for him. That 
complicated an already difficult situation because 
those who took over his tasks favored a much 
simpler machine than the SIS was expecting. 75 

TSKThere were meetings, but they were disap- 
pointing. The NBS was willing to promise a 
design for only a very bare bones device. And they 
could not guarantee when those plans would be 
ready for the engineers and programmers who 
waited for them at Arlington Hall. 

(S//SI) Everyone at the SIS grew more frus- 
trated. The frustration was compounded by grow- 
ing ambitions. As the SIS engineers and pro- 
grammers gained more experience, they thought 
of many ways to make "Abner" an effective crypt- 
analytic device. 76 But it would have to be a much 
more complex machine than the one proposed by 
the NBS and a more intricate one than Atlas. It 

was also clear that Abner was going to cost the 
agency much more than had been imagined. 

(U) So Much for Simplicity 

ITSffSi^'With young men like Ray Bowman 
in the lead, ideas were put forward to change 
Abner into something more like a Sled than a 
simple EDVAC. He and others showed how basic 
cryptanalytic functions could be turned into cir- 
cuits that, they thought, should become an inte- 
gral part of Abner. Some fifteen special "instruc- 
tions" were drawn as circuit diagrams and were 
shown to the few men at the NBS who had secu- 
rity clearances. 77 Their reaction was not positive; 
they felt they were being asked to do much more 
than was initially agreed upon. And even when a 
compromise was suggested, that the special func- 
tions such as a two-message offset instruction be 
put into a separate box that the SIS would design 
and build, agreements could not be reached. Even 
such a box, the NBS engineers argued, would 
demand too many complex circuits in the main 
computer. 78 There was a stand-off, a quiet one, 
but it was clear the two agencies had reached an 
impasse in late 1949. 79 

(U) Another critical decision had to be made. 
Should Abner be abandoned at least until the 
uncontrollable NBS decided to devote serious 
attention to it? Or, since the SIS and OP-20-G 
had been merged into the new Armed Forces 
Security Agency, should everyone be required to 
wait for the completion of the further advanced 
Atlas project? It did not take too long for the 
administrators to make a courageous decision. 
They allowed the Arlington Hall engineers to go 
ahead and design and build their own machine. 

(U) That seemed a reasonable decision 
because so much agency effort had already gone 
into Abner and because the computer manufac- 
turers, including IBM, continued to back away 
from taking contracts for machines. And the 
agency knew that smaller companies, such as 
Technitrol, could be counted on for components. 


Page 253 


They were already helping with the special-pur- 
pose machines the agency was considering. 

(U) Abner's Not Quite Best Friend 

f©-The decision to build Abner received an 
unexpected justification in mid-1950. During the 
first months ofthe Korean War, when the 
American military was unable to show that it 
could effectively police the world, it was at least 
suspected that the American navy's callsigns were 
inadequate. New ones had to be generated to pro- 
tect the fleet and its messages. There was a criti- 
cal need to run "involuntary matrices" to ensure 
randomness. That was a demanding job and one 
that Solomon Kullback, the agency's research 
director, had given the highest priority. 

4$) In summer 1950 a survey was made ofthe 
possibility of running the matrices on IBM tabu- 
lator equipment. The result was very disappoint- 
ing. Then a suggestion was made that perhaps the 
relay analog ofthe soon-to-be-delivered Atlas 
could be used. Some ofthe mathematicians 
explored the possibility and then, perhaps 
prompted by Sam Snyder, looked at the possibil- 
ity of using the NBS's new SEAC. It was the inter- 
im computer the NBS had decided to build on its 
own in 1948. Constructed in two years, it was a 
bare bones version ofthe EDVAC, but it worked. 

JJ&) Required to estimate whether or not it 
would be worthwhile to ask the NBS if SEAC 
could be used on the callsign problem, the SIS's 
programmers made pencil and paper calculations 
ofthe power of various alternatives. 

•&) Their estimates gave the following times 
for gaining a completed and satisfactory matrix. 80 


By Hand By Relay ByArJ 

5 hours 4.5 hours 3 minutes 

(S^ The results seemed unambiguous. The 
electronic machines were undoubtedly faster 

By Seac 
20 seconds 

than hand or relay process, and most important- 
ly, the delay-line machines, such as SEAC, were 
orders faster than the ones with drum memory. 
That finding was support for Abner's cause and 
for the NBS's SEAC. As a result, Sam Snyder went 
to the National Bureau of Standards and gained 
permission to use the SEAC for the matrix prob- 
lem. 81 Unfortunately, the first hands-on experi- 
ence with an electronic computer was very disap- 
pointing. If there had not been so much invested 
in Abner, the experience with SEAC might have 
led to abandoning Abner. 

-QS) A program for computing the matrices on 
SEAC was written in some two months, not an 
easy task in the early 1950s. Then in September 
some SEAC run time was allocated to the agency 
- but on weekends and nights and at $24.00 an 
hour, not an inconsiderable sum at the time. The 
expense had not been foreseen, nor had the 
SEAC's temper. SEAC soon taught the analysts 
that there was a vast difference between the theo- 
retical internal speed of computers and their real 
operating time. 

C&) The first post-midnight session on SEAC 
took twelve very discouraging hours. Despite all 
the care that had been taken, the result had to be 
abandoned because of repeated machine errors. 
A few days later some 200 matrices were created, 
but then errors crept in again. After just two 
hours of successful runs, it was decided that a 
checking program should be written so that all 
results could be verified. 

^g) The frustrations continued. In mid- 
September, SEAC worked only four out of sixteen 
hours, and the results that were obtained were 
put into question because ofthe quirky behavior 
ofthe computer. 

S&) On September 19 SEAC did more than fol- 
low the laws of early computer technology (to 
rarely work); it committed a serious political 
blunder. The agency had reserved a precious 
twelve-hour chunk of time, between noon and 

Page 254 


rop msmmasmmns: usa, auo, can odr and i mm 

midnight, because some official had decided it 
was appropriate that the great secret of SEAC and 
the future Abner be revealed to the SIGINT com- 
munity. Some twenty-five people were invited to 
watch the NBS's machine in action. Sam Snyder 
was in charge of running SEAC. 

(S) Much to Snyder's embarrassment, of the 
entire twelve hours "o hours were productive." 
He was so angry that he wrote in a report, "In the 
future, when trouble with SEAC develops, no 
more attempts will be made by personnel of this 
Section to find the cause of the difficulty." 82 

(S) Despite the anger and the problems that 
continued to be encountered with SEAC runs, it 
was decided to keep the matrices that had been 

(U) Abner by Inertia 

f€£. While SEAC was refusing to give its best, 
the crew at Arlington Hall got to work building 
Abner, not telling the NBS exactly what they were 

doing. They worked so quickly and quietly that 
when the NBS representatives finally appeared at 
Arlington Hall with some sketches, they were 
shown, with much ceremony, the completed SIS 
design. 83 

{6} That design had become very ambitious, 
going far beyond the NBS's SEAC. 84 The basic 
EDVAC approach was maintained, but many spe- 
cial functions were included. Abner became much 
more of a crypto-computer than Atlas. 

69) Three groups of special instructions were 
built into Abner. The first made encipher/deci- 
pher tasks easier. A programmer could call for 
addition without carrying and specify which 
number base, ranging from two to greater than 
thirty-two. Abner could very quickly run key 
against text, imitating many of the previous spe- 
cial-purpose machines. The second group of com- 
mands made Abner a more efficient processor of 
streams of data, allowing easy transfer of blocks 
of information and repetition of a series of 
instructions until a task was completed. The third 

•4^ Abner 

top or.onnv/oo M!N T//m:L uoa, auo, can odr a n d n zujx i 

Page 255 


group contained instructions that made Abner a 
powerful "comparator." Using the one Swish 
instruction, a programmer could tell Abner to 85 

Pass two variable length streams of five-bit 
characters from memory to the control of the ana- 
lytic unit of the machine; 

Compare groups of varied sizes (one to sixty- 
three characters) for coincidences; 

Store the count in a specified location; 

Offset one data stream to prepare it for anoth- 
er round of coincidence testing. 

•fS} As significant for the evolution of the com- 
puter, there were courageous attempts to give 
Abner what most other computers of the time did 
not have, a range of powerful input and output 
devices. A big gamble was taken: the new plastic 
magnetic tapes could be made to function. 
Connections for six of the Raytheon drives were 
installed. A punched tape reader was attached as 
well. Not as exciting, but more important from an 
operational standpoint, an IBM collator was to be 
used for card input and a modified IBM card 
punch for output. 86 There were many software 
developments. By the time Abner was completed 
in April 1952, the SIS programmers had written a 
wide range of operational routines. 87 

(U) Abner's Bad Temper 

fjS)-But it was some two years between the 
time the SIS engineers decided to detail their own 
Abner and its start-up as an operational machine. 
And its cost climbed to twice the original estimate 
of $300,000. Even then, it was just, as one engi- 
neer put it, an "experimental model." 

-(S) And it was almost as temperamental as 
SEAC. Its special functions made it a bit too com- 
plicated to maintain (it had 1,500 tubes and 
25,000 diodes), the PO devices and their inter- 
faces had many troublesome moments, and the 

more than 100 delay lines needed constant fine- 


""($)• The limitations of the 1952 Abner were 
obvious. But its problems did not cause the SIS to 
abandon computers. A contract was let to have 
the Technitrol company in Philadelphia build a 
new version. It arrived in mid- 1955, cost approx- 
imately $1,000,000, and like Abner I, had an 
operational price tag of almost the same amount. 

-6S-) It was known in the 1940s that operators, 
cooling systems and repairs would make comput- 
ers very expensive to maintain. But there was 
another cost that was not anticipated: program- 
ming. Abner I needed $130,000 a year worth of 
programmer time plus additional amounts for 
special projects. 89 

"XS//SB Abner had another expense: a clone. 
In 1950 it was decided that Abner should have the 
same kind of relay-circuit cousin that had been 
quickly built to train Atlas's programmers. When 
the construction of "Baker" began, no one expect- 
ed it to take two years to complete, nor to become 
the size of a room. Nor did anyone foresee that 
the relay version of Abner would be much less 
reliable than the new electronic machine. Baker 
proved so difficult that it never kept its promise to 
be an inexpensive training and program-debug- 
ging aid to Abner. 90 

4S) Despite Baker's failings, many innovative 
programs were written for Abner. They spanned 
all cryptanalytic attacks as well as data processing 
tasks. The list of programs is impressive, espe- 
cially when it is realized they were written in an 
era when programming was something of a black 
art. For many years programs had to be written in 
the o's and l's the computer recognized, and even 
when "higher" level languages appeared, pro- 
gramming and debugging were energy-draining 
and emotion-laden exercises. 

£KS) Abner was used to "diarize" as well as to 
analyze wired rotor systems. But perhaps its most 

Page 256 



useful program was Stethoscope. Written very 
early in the history of programming, it became a 
classic. Stethoscope very efficiently applied all the 
major statistical attacks against cipher text in 
unknown systems. 91 The routines proved so valu- 
able, and the potentials of expanding Stethoscope 
seemed so great, that a very courageous step was 
taken by the SIS programmers. They decided to 
write one of the very first compilers. Bill Cherry 
played a key role in the LULU project to create 
software that would allow programmers to very 
easily compose and correct programs that called 
many subroutines. With the help of LULU and 
its follow-on, Stethoscope became the much 
more powerful general cryptanalytic program, 

(U) And Then Came ... 

(fS/ySE) It had taken many years and millions 
of dollars to prove that a universal computer 
could be a valuable statistical tool, however. And 
there were some critical moments when it seemed 
that a special architecture would be the wisest 
choice for the SIGINT agencies. In fact, one of the 
gravest crises in the history of American crypt- 
analysis shifted attention to such alternatives and 
highlighted the weaknesses of the new general- 
purpose computers. "Black Friday" of 1948 saw a 
return to a faith in special-purpose devices; they 
seemed the only way to overcome Soviet systems 
that were making the Enigma and even Tunny 
look simple. But "Black Friday" also showed how 
far the computer had to go before it could replace 
the old reliable data processors, the tabulators. 


i.(U) NSA, OP-20-G, J.T. Pendergrass, 
"Cryptanalytic Use of High-Speed Digital Computing 
Machines," 1946. NSA, Samuel S. Snyder "Influence of 
United States Cryptologic Organizations on the Digital 
Computer Industry," dates the beginning of Goldberg 
to 1947 perhaps on the basis of the contract for the 
machine rather than on the date of start of the explo- 
rations for a universal scanning machine. 

2. (U) Interview with Philip J. Bochicchio, July 
1994, On Rachman, (T0//0I) NSA CCH Series XII Z, 
and CCH Computer History Box, OP-20-G "War Diary 
Reports: March 1, 1943 - May 31, 1948," August 1945. 

3. (U) Anthony Ralston (ed.), Encyclopedia of 
Computer Science (New York: Van Norstrand 
Reinhold, 1976), 482, 1459. 

4-fe)-NSA CCH Series XII Zand XI K, Snyder, 
Box 9, Samuel S. Snyder, "And Then There Were Two, 
the Abner Story," Fourth Draft, December 1979, 1-2. 

5. (U) On the postwar Selector project and its tie to 
Bush and "G's" ERA, see Colin Burke, Information 
and Secrecy: Vannevar Bush, Ultra, and the Other 
Memex (Metuchen, N. J.: The Scarecrow Press, 1994.) 
On Bush and Eisenhower, see Thomas Johnson, 
American Cryptology during the Cold War: Book I, 
Ft. George Meade, MD: NSA Center for Cryptologic 
History, 1995. 

6. (U) Colin Burke, Information and Secrecy: 
Vannevar Bush, Ultra, and the Other Memex, 
(Metuchen, N. J.: The Scarecrow Press, 1994), Ch. 14. 

7.(U) NSA OP-20-G, J.T. Pendergrass, 
"Cryptanalytic Use of High-Speed Digital Computing 
Machines," 1946. (U) NSA, Samuel S. Snyder, 
"Influence of United States Cryptologic Organizations 
on the Digital Computer Industry" dates the beginning 
of Goldberg to 1947, perhaps on the basis of the date of 
the contract for the machine rather than on the date of 
the start of the explorations for a universal scanning 

8.(IS#ei) NSA CCH Series XII Z,H. H. 
Campaigne, "Reading TUNNY," NSA Technical 
Journal, (Fall 1962). ( J8//S BJ A fascinating source for 
the history of OP-20-G are Campaigne's War Diaries, 


9. (U) NSA CCH Series XI K, Snyder, Box 8, 
OP-20-G4, "Report on conference held at Navy 
Department 15 May 1946. "Survey of large scale auto- 
matic computing machines, given by J. von 
Neumann," Howard Campaigne, 16 May 1946. 

10. (U) Martin Campbell-Kelly and Michael R, 
Williams (ed.), The Moore School Lectures: Theory 
and Techniques for the Design of Electronic Digital 
Computers (Cambridge, Mass.: MIT Press, 1985). 

11. (¥8} NSA CCH Series XII Z, NSASAB, 
"Historical Study, Lightning-Freehand," circa 1963. 


Page 257 


(TO//6I) NSA CCH Series XII Z, "Mechanization in 
Support ofCOMINT, Phase 111: Third Addition," i 
November 1956. (U) "Remarks at the Dedication of 
John von Neumann Hall/' NSA Technical Jourrad, VI 
(Winter 1961): i-CB) NSA CCH Series XII Z, Morns 
Pomerantz and Lawrence A,Sames, "Data 
Distribution Network for the TABLON Mass Storage 
System," circa 1970. 

12. fPS) NSA CCH Series XII Z,Lt. Cdr. J.T. 
Pendergrass, "High Speed Digital Computing 
Machines, Cryptanalytic Uses of," 15 October, 1946- 
( T0//0I) NSA CCH Series XII Z, H. H. Campaigne and 
J.T. Pendergrass, "Second Report on Cryptanalytic 
Use of High Speed Digital Computing Machines," 0P- 
20-L, 18 December 1946. Campaigne wrote one of the 
programs in the first report, but that has not been 
remembered. He was listed as the joint author in the 
second one 

13. (U) It is important to note that the reports did 
not claim that the machine was an efficient data 
processor. All of the many examples were cryptanalyt- 
ic ones, and processing-heavy methods such asT/A 
were not mentioned. 

14. PES) NSA CCH Series XII Z,Lt. Cdr. J.T. 
Pendergrass, "High Speed Digital Computing 
Machines, Cryptanalytic Uses of," 15 October 1946, 1. 

15. (U) von Neumann had, at first, favored a four- 
address system, then changed to the one-address idea 
because it would be better suited to mathematical 
machines. Pendergrass thought that the one-address 
was best for crypto-work. In contrast the SIS comput- 
er pioneers favored the four-address system. 

16. PT0//3 -T ) NSA CCH Series XII Z, H. H. 
Campaigne and J.T. Pendergrass, "Second Report on 
Cryptanalytic Use of High Speed Digital Computing 
Machines," OP-20-L, 18 December 1946. The available 
copy of this report did not contain the Mercury pro- 
gram, but it was cited as part of the report on page 1. 

17. (M NSA AHA ACC 32685, folder, "Atlas 
Computer Correspondence," "Status of Digital 
Computers, November 1946." 

18. C*y NSA AHA ACC 32685, folder, "Atlas 
Computer Correspondence," "Status of Digital 
Computers, 20 January 1947." 

19. ("R^He included Alan Turing's work. 

20.t$-NSACCH Series XII Z, Ann M.Ford, "The 
Birth of Atlas 1, NSA Technical Journal, XVIII (Winter 

1973) ; 53- 

21. GHWST) NSA CCH Series XII Z, "File Kept by 

Dr. Campaigne on RAM Panel Meetings," notes on the 
13 November 1946 meeting clearly show that the digi- 
tal computer was seen as a practical cryptanalytic tool, 
not just a research machine. 

22. m NSA AHA ACC-32685, folder, "Atlas 
Computer Correspondence," "Memorandum for 
Research Division and Section Heads," 23 January 


23. m NSA CCH Series XII Z, Ann M. Ford, The 
Birth of Atlas I," NSA Technical Journal, XVII [ 
(Winter 1973); 53- 

24. (U) On the interesting Aiken designs, Michael 
K. Williams, A History of Computing Technology, 
(Englewood Cliffs, N.J: Prentice-Hall, 1985). 

25. ^S) NSA AHA ACC 32685, folder, "Atlas 
Computer Correspondence," "Estimated Delivery to 
NCML," circa 1951. Atlas was given anAA priority; 
that was higher than Goldberg's, at least during the 
late 1940s. 

26. m NSA AHA ACC 32685, folder, "Atlas 
Computer Correspondence," CNO to Chief of Bureau 
of Ships, 23 May 1947." 

27. ffS#ST) NSA CCH Series XII Z, "File Kept by 
Dr. Campaigne on Ram Panel Meetings," notes of 
meeting of 22 November 1946, page 2, shows "G" 
believed RCA and the NBS were soon going to build a 

28. pFfl//3I) NSA CCH Series XII Z, 
"Communications Supplementary Activities, RAMP 
Report II," 21 December 1948, 10. ( ¥S//OI j NSA CCH 
Series XII Z,"File Kept by Dr. Campaigne on Ram 
Panel Meetings," report of 14 November 1946 meeting. 

29. (Q NSA CCH XI K, Box 8, Snyder, "An 
Evaluation ofNSA's Atlas I," Ann Ford H12, 8 
November 1970. March seems to have been a critical 
month in Atlas history. "G" decided to build its own 
machine because no vendor seemed willing to do so. 

30. ($ NSA AHA ACC 32685, folder,"Atlas 
Computer Correspondence," "Summary of Conference 
on Task 13 (Atlas)," 19 and 21 August 1947. 

31. (§) NSA AHA ACC 32685, folder, "Atlas 
Computer Correspondence," "Second Revision of 

Page 258 


t o p nconcT/zo oMi HTOn r i uoa, auj, law opn mhu m u/x i 


Military Characteristics of the Analytic Computi 
(ATLAS)," 6 July 1948. 

32. (U) W. W. Stifler (ed.), High-Speed Computing 
Devices, (New York: McGraw-Hill Book Company, 
1950), 370. 

33- (U) Apparently, the proposed MIT tube was 
quite like Wilkes'. 

34- (U) The result, a Selectron, was not available 
until the early 1950s. It was used in one von Neumann 
type of computer, but never afterward. One of the 
many leaps in computer technology had made it a 
technological dinosaur. 

35- (•¥») NSA CCH Series XII Z, file folder, 
"Monogram and RAM Panel Reports, 1945-1949," 
Meeting of April 15, 1947. 

36. ( TD//0 E ) NSA CCH Series XII Z, 
"Communications Supplementary Activities, RAMP 
Report II," 21 December 1948. Electrostatic storage 
had also been planned for Goldberg. 

37. ( TS//S *) NSA AHAACC 11112, "Interim Report 
on Computer Research," circa 1948. 

38. (U) On microfilm, (S) NSA CCH Series XI K, 
Sam Snyder, Box 12, "Analytic Machinery Principles," 
September 1949. On Atlas' drum, ^¥8)"NS A AHA ACC 
13643 "Atlas I. " 

39. (U) Philip J. Bochicchio has recounted his 
experiences with the earliest magnetic drums. He stat- 
ed that in 1945, after gaining access to captured 
German equipment, he created a primitive drum that 
was taken by Joe Eachus to ERA That inspired ERA 
stated Mr. Bochicchio. 

40. (¥S) NSA CCH Series XII Z, "Report of the 
Second Computer Study Group," as in NSA Technical 
Journal, XIX (Winter 1974): 21-61. (TS) NSA CCH 
Series XII Z, file folder, "Monogram and RAM Panel 
Reports, 1945-1949." 

41. {S} There were problems with the Goldberg 
drums, and they were run at a low 240 rpm. (S) NSA 
CCH Series XII Z, "Goldberg Progress Reports," 30 
December 1947 through 10 April 1951. 

42. (TS//SI ) NSA CCH Series XII Z, 
"Communications Supplementary Activities, RAMP 
Report II," 21 December 1948. (SS-) NSA AHA 36746, 
Engineering Research Associates, Inc., "Proposal 
for An Electronic Rotor Program," 19 December 
1946. (IS.) NSA AHA ACC 8252, 0P-20-G, 

"Communications Intelligence Research Plans 1948 " 
7April 1947. mffm) NSA CCH Series XII Z, "Joint 
Long Term Program (Old Planning Material, 1948- 
1949) compiled by Doug Hogan.tS^ NSA AHA ACC 
32685 "Summary of Conference on Task 13 (Atlas)," 19 
and 21 August 1947. 

43- 66} J.J. Eachus, "SIGMAGE Threshold 
Control," 2 July 1946. (¥S» NSA CCH Series XII Z, file 
folder, "Monogram and RAM Panel Reports, 1945- 
1949." Monogram's proposed allocations for computer 
research for 1949 were quite generous: $1,000,000 for 
general-purpose computer work and $1,000,000 for 
SPD and related electronic work. (§} NSA CCH Series 
XII Z,BuShips, "Specifications Sled Navy Models 
CXOA and CXNQ Block Diagrams," 1 October 1948, 
Monogram's 1948 budget was cut in half by the 
bureau, (¥S) NSA CCH Series XII Z,file folder, 
"Monogram and RAM Panel Reports, 1945-1949," 
entry for 21 November 1947. 

44- (TS//SI) NSA CCH Series XII Z, "File Kept by 
Dr. Campaigne on Ram Panel Meetings," notes on 27 
July 1948 meeting. CS)-NSA AHA ACC 32685, folder, 
"Atlas Computer Correspondence," "Estimated 
Delivery toNCML," circa 1951. (U) Hagley Museum 
and Library, Accession 1901, Yuter Papers, May 20, 
1947, ERA Tompkins Report on Atlas, "shift Goldberg- 
Demon men to project." (U) Erwin Tomash, "The Start 
of an ERA: Engineering Research Associates, Inc., 
1946-1955," in N. Metropolis, et al., (ed.), A History of 
Computing in the Twentieth Century (New York: 
Academic Press, 1980), 485-496. 

45. CS) NSA Technical Literature Series, 
Monograph No. 2, History of NSA General-Purpose 
Electronic Digital Computers, 1964, 9. 

46. (6) NSA CCH XI K, Box 8, Snyder, "An 
Evaluation ofNSA's Atlas I," Ann Ford H12, 8 
November 1970. 

47. (C^-NSA AHA ACC 32685 "First Endorsement 

on ERA Inc 23 February 1951. (U) Samuel S. 

Snyder "Influence of United States Cryptologic 
Organizations on the Digital Computer Industry," The 
Journal of Systems and Software, 1 (1979): 90-91. 
(U) Hagley Museum and Library, Accession 1901, 
Yuter Papers: Engineering Research Associates, 
October 9, 1946, Meeting on NBS computer plans, 
"Summary of Computing Conferences"; Tompkins to 


Page 259 


Norris, October 19, 1946, "Computing Business"; 
December 1946, "Reports onOP-20-G Projects and 
Atlas Computer"; and Goldberg Report, June 27, 1947. 

48.tS7YSI) NSA CCH Series XII Z, NSA Technical 
Literature Series, Monograph No. 2, History of NSA 
General-Purpose Electronic Digital Computers, by 
Samuel S. Snyder, 1964, 7- t&r NSA AHA ACC 32685, 
"Atlas I" circa 1952. 

49- (9 //BQ NSA CCH Series XII Z, NSA Technical 
Literature Series, Monograph No. 2, History of NSA 
General-Purpose Electronic Digital Computers, by 
Samuel S. Snyder, 1964, 7. 

50. (¥S) NSA AHA 6851 Atlas Analytic Computer, 
"Military Characteristics of the Analytic Computer 
(Atlas)," June 1947- NSA AHA ACC 13643, 
"Memoranda on Electronic Computers, Atlas I," circa 
1952. (S^-NSA AHA ACC 32685 "AFSA-351 Atlas 
Programming Bulletin No 1," February 1951. 

51. te)-NSA CCH XI K, Box 8, Snyder, "An 
Evaluation ofNSA's Atlas I," Ann Ford H12, 8 
November 1970, 8. 

52. (8 //0r ) NSA CCH Series XII Z, NSA Technical 
Literature Series, Monograph No. 2, History of NSA 
General-Purpose Electronic Digital Computers, by 
Samuel S. Snyder, 1964, 8. 

53-£S} NSA CCH Series XII Z, "Report of the 
Second Computer Study Group," as in NSA Technical 
Journal XIX (Winter 1974): 21-61. *Q NSA CCH 
XI K, Box 8, Snyder, "An Evaluation ofNSA's Atlas I," 
Ann Ford H12, 8 November 1970. 

54- &) NSA AHA ACC 32685, folder, "Atlas 
Computer Correspondence," CNO to Chief of Bureau 
of Ships, 23 May 1947. 

55- {SfNSA CCH Series XI K Box 8, Snyder, 
"Yearly Cost of Representative NSA Machines," May 


56. {S} NSA Technical Literature Series 
Monograph No. 2, History of NSA General-Purpose 
Electronic Digital Computers, 1964, 11. (TS//SI) NSA 
AHA ACC 30851, "Historical Notes on Computers at 
NSA" suggests that metallic tapes were planned. 

57. (U) On the naming of the machine, (S //3 i) 
NSA CCH Series XII Z, Oral History Interview OH 04- 
82 with Samuel S. Snyder, 24 February 1982. The most 
complete survey of Abner is (6)-NSA CCH Series XII Z 
and XI K Snyder, Box 9, Samuel S. Snyder, "And Then 

There Were Two, the Abner Story," Fourth Draft, 
December 1979. See also, (U) Samuel S.Snyder, 
"Abner: The ASA Computer, Part I: Design," NSA 
Technical Journal,XXV No. 2 (Spring, 1980): 49. (U) 
Samuel S. Snyder, History of NSA General Purpose 
Electronic Digital Computers , 1964. 

58. (U) The electronic group had some sixty 
employees in 1946. 

59. (C) NSA CCH Series XII Zand XI K Snyder, 
Box 9, Samuel S, Snyder, "And Then There Were Two, 
the Abner Story," Fourth Draft, December 1979, 1-6. 

60. Wf&) NSA CCH Series XII Z, Oral History 
Interview OH 04-82 with Samuel S. Snyder, 24 
February 1982, 70. 

61. t^-NSA CCH Series XII Zand XI K Snyder, 
Box 9, Samuel S. Snyder, "And Then There Were Two, 
the Abner Story," Fourth Draft, December 1979, I-10. 
(¥9) NSA CCH Series XI K, S. Snyder, Box 8, Mary 
Neely Roseboro, CSGAS-76C, "Commentary on the 
Pendergrass Report," 15 October 1947. 

62. £S#Sf) NSA CCH Series XII Z, Oral History 
Interview OH 04-82 with Samuel S. Snyder, 24 
February 1982, 121. 

63. (U) See also, R. M. Bloch et al., "Logical Design 
of the Raytheon Computer," Mathematical Tables and 
Other Aids to Computation, 3 (October 1948): 286. 

64. (S//6I) NSA CCH Series XII Z, Oral History 
Interview OH 04-82 with Samuel S. Snyder, 24 
February 1982, 121. The UNIVAC design had a serious 
shortcoming for cryptanalytic work. The standard 
UNIVAC had a decimal (BCD) organization. That 
meant it was not suited to handle many crypto-tasks. 
Especially important were the new targets of the cryp- 
to-groups, the baudot-based binary systems. The 
analysis of such systems demanded bit-by-bit testing, 
as did many of the older targets. ( fi//3 f) NSA CCH 
Series XII Z, NSA Technical Literature Series, 
Monograph No. 2, History of NSA General-Purpose 
Electivnic Digital Computers, by Samuel S.Snyder, 
1964, 13- 

65. CflS} NSA CCH Series XII X-MPRO, U.S. 
Cryptanalytic Research and Development Committee, 
"Joint Long Term Program for Research and 
Development in the Field of Cryptanalytic 
Equipment," 21 July 1948. (S) NSA CCH Series XI K 

Page 260 



Box 8, Sam Snyder, "Evaluation of Computers as Crypt 
Aids," 7 September 1948. 

66. (TTT Mina Rees, "The Mathematical Sciences 
and World War II, American Mathematical Monthly, 
87(1980): 607-621. William Aspray and Michael 
Gunderloy, "Early Computing and Numerical Analysis 
at the National Bureau of Standards," Annals of the 
History of Computing ,11 (1989): 3-11. John Todd, 
"John Hamilton Curtiss, 1909-1977," Annak of the 
History of Computings (1980): 104-9. R - Cochrane, 
Measures for Progress: A History of the National 
Bureau of Standards (Washington: G. P. O., 1966). 
Samuel S.Snyder, "Abner: The ASA Computer, Part 1: 
Design," NSA Technical Journal, 25 (1980): 49. 

67. ($ffS¥) NSA CCH Series XII Z, NSA Technical 
Literature Series, Monograph No. 2, History of NSA 
General-Purpose Electronic Digital Computers, by 
Samuel S. Snyder, 1964, 13.{€)-NSA CCH Series XII Z 
and XI K Snyder, Box 9, Samuel S. Snyder, "And Then 
There Were Two, the Abner Story," Fourth Draft, 
December 1979, 1-13. 

68. {G) NSA CCH Series XII Z and XI K Snyder, 
Box 9, Samuel S, Snyder, "And Then There Were Two, 
the Abner Story," Fourth Draft, December 1979, 1- 13 

69. (3//3I) NSA CCH Series XII Z, NSA Technical 
Literature Series, Monograph No. 2, History of NSA 
General-Purpose Electronic Digital Computers, by 
Samuel S.Snyder, 1964, 12. Raytheon also chose a 
more data-oriented design with four addresses for its 

70. (¥8} NSA CCH Series XI, Snyder, Box 8, 
folder, "Snyder Computer Trips, 1947-1951." (€} NSA 
CCH Series XII Zand XI K Snyder, Box 9, Samuel S. 
Snyder, "And Then There Were Two, the Abner Story," 
Fourth Draft, December 1979, 1-14. 

71. (U) James W. Cortada, Historical Dictionary 
of Data Processing: Biographies (New York: 
Greenwood Press, 1987), 64. 

72. (U) NSA CCH Series XII Zand XI K Snyder, 
Box 9, Samuel S. Snyder, "And Then There Were Two, 
the Abner Story," Fourth Draft, December 1979, 1-15. 

73- (U) Samuel S.Snyder, "Abner: The ASA 
Computer, Part I: Design," NSA Technical Journal, 25 
(1980): 49. Samuel S.Snyder, History of NSA 
General-Purpose Electronic Digital Computers , 1964, 
15. {GfNSA CCH Series XII Z and XI K Snyder, Box 9, 

Samuel S, Snyder, "And Then There Were Two, the 
Abner Story," Fourth Draft, December 1979, II-3. 

74-"^ NSA CCH Series XII Z, H. F. Gingerich, R. 
A. Leibler, "Hagelin Crib D Dragging on a High Speed 
Automatic Computing Machine," 26 August 1949. (£}• 
NSA CCH Series XI K Box 8, Sam Snyder, "Evaluation 
of Computers as Crypt Aids," 7 September 1948. 

75. (U) Samuel S.Snyder, "Influence of United 
States Cryptologic Organizations on the Digital 
Computer Industry," The Journal of Systems and 
Software, 1 (1979): 92. 

76. fS^S*) NSA CCH Series XII Z, NSA Technical 
Literature Series, Monograph No. 2, History of NSA 
General-Purpose Electronic Digital Computers by 
Samuel S. Snyder, 1964, 14. 

77- ($&&) NSA CCH Series XII Z, NSA Technical 
Literature Series, Monograph No. 2, History of NSA 
Geneml-Purpose Electivnic Digital Computers , by 
Samuel S. Snyder, 1964, 15, gives a full list of the spe- 
cial functions including the "Swish," which was "the 
logical equivalent of a complete high-speed compara- 
tor." The proposed special function of Abner found its 
way into the later Harvest machine. 

78. (¥S) NSA AHA ACC 10842, Ray L. Bowman, 
"Engineering Diary," circa 1945-1950. (S//SI) NSA 
CCH Series XII Z, Oral History Interview OH 04-82 
with Samuel S. Snyder, 24 February 1982. 

79--^5 NSA CCH Series XII Zand XI K Snyder, 
Box 9, Samuel S. Snyder, "And Then There Were Two, 
the Abner Story," Fourth Draft, December 1979, II- 15. 
The security clearance problem seems to have, again, 
complicated matters. It made communications diffi- 
cult. The NBS faced severe problems because of "loyal- 
ty" questions in the early 1950s. 

80. rSj NSA CCH Series XII Z,AFSA-32, 
Marvine Bass, "On Methods and Speed of 
Construction of Involuntary Matrices," August 1950. 

81. (S) NSA CCH Series XI K, Snyder Box 10, 
"Extracts from AFSA-351D Weekly Reports reSeac 

82. (&) NSA CCH Series XI K, Snyder Box 10, 
"Extracts from AFSA-351D Weekly Reports re Seac 

83. ^ NSA CCH Series XII Zand XI K Snyder, 
Box 9, Samuel S. Snyder, "And Then There Were Two, 
the Abner Story," Fourth Draft, December 1979, III-19. 


Page 261 


84. OS) NSA CCH Series XII Z and XI K Snyder, Box 
9, Samuel S. Snyder, "And Then There Were Two, the 
Abner Story," Fourth Draft, December 1979, 1-2. 

85. -ffi^-NSA Technical literature Series, Monograph 
No. 2, Samuel S.Snyder, History of NSA General- 
Purpose Electronic Digital Computers, 1964, 16. 

86. fS) A useful personal insight into Abner is 
Russell Chauvenet, "Early Days in NSA Computing," 
Cryptolog , August 1977: 8-10. 

87. (S}-NSA Technical Literature Series Monograph 
No. 2, Samuel S.Snyder, History of NSA General- 
Purpose Electronic Digital Computers, 1964, 17. Abner 
was ready for its checkout phase in September 1951. All 
its instructions were accepted in April 1952. 

88. (€) NSA CCH Series XII Zand XI K Snyder, 
Box 9, Samuel S. Snyder, "And Then There Were Two, 
the Abner Story," Fourth Draft, December 1979, IV-14. 
(U) NSA CCH Series XII Z,to Cryptolog editor, by 
R. L. Bernard, "Comments on Abner," 18 January 1978. 

89. &)■ NSA CCH Series XI K Box 8, Snyder, "Yearly 
Cost of Representative NSA Machines," May 1955. 

90. (3//3I) NSA CCH Series XII Z, NSA Technical 
Literature Series, Monograph No. 2, History of NSA 
General-Purpose Electronic Digital Computers, by 
Samuel S. Snyder, 1964, 18. 

91. <£)- NSA CCH Series XII Z and XI K Snyder, 
Box 9, Samuel S. Snyder, "And Then There Were Two, 
the Abner Story," Fourth Draft, December 1979, IV-8. 
(TS) NSA CCH Series XI K, Snyder, Box 16, List of 
Operational Abner Programs. 



Chapter 9 
(U) Wandering into Trouble 

(U) A Cryptanalytic Future 

TflS) Although OP-20-G and the SIS were 
frustrated by the delays in their Atlas and Abner 
computer projects, 1948 began as a year of con- 
tinued triumph for American communications 
intelligence. The cryptologic systems of the major 
powers were being read, and COMENT seemed 
about to supply America's leaders with the type of 
high-level information that had won "G" and the 
SIS so many accolades in World War II. 

Q¥&) Most of Stanford C. Hooper's dreams 
seemed to have been realized. The army and navy 
COMINT organizations had some professional 
mathematicians, they were starting vibrant new 
RAM programs, and they even had "scientific" 
advisory boards. The future of advanced crypt- 
analysis was bright, as was the future of American 
intelligence in general. 

(¥&} Most important, the Cold War, it 
seemed, was to be a cryptanalytic one. A few for- 
eign systems carrying high-level messages would 
yield to mathematics and computers and provide 
the kind of information needed to predict and, 
perhaps, counter the actions of the political and 
military leaders of all the important nations. 
There was even cryptanalytic progress against 
internal enemies. Old diplomatic messages were 
giving what was needed to find and break up 
Soviet espionage rings within the United States. 
There was hope that further works would lead to 
entries into all current Eastern bloc diplomatic 
and clandestine systems. 

TO Cryptanalytic success would not be 
expensive. Direction finding as well as the very 
labor-intensive traffic and plain language analy- 
ses would play secondary roles. America's 
COMINT agencies would not have to build a 

costly communications system to speed massive 
amounts of data to processing centers because 
high-level cryptanalysis provided lead time. 
Cracking important systems would give insight 
into the grand intentions of the world's political 
and military leaders, and that would give 
American policy makers time to formulate meas- 
ured responses. 

TTSD The belief that cryptanalysis would 
be the heart of SIGINTs future was reflected in 
the kind of machines the army and navy devel- 
oped during the immediate postwar years. They 
focused on general-purpose computers for crypt- 
analysis, not ones for massive data processing. 

(U) The Worst of Times 

(TS//S *j Then the heroic cryptanalytic 
assumption was suddenly undermined! In spring 
1948 the SIGINT agencies had to rethink their 
purpose and place in America's intelligence 
establishment. The Americans and their intelli- 
gence partner, the British, were being locked out 
of the world's most important code and cipher 
systems just as the Cold War became dangerous. 
Three years after the end of the war, the Soviets 
closed down their old high-level systems and 
replaced them with ciphers that could not be pen- 
etrated. Soon the Chinese revised their superen- 
cryption s, making them quite sturdy. Even the 
strengthened their machines and proce- 


CTS//SIj A series of very significant intelli- 
gence failures came after the cryptanalytic black- 
out. That put American SIGINTs future in jeop- 
ardy. The Soviets' A-Bomb, the Berlin Blockade, 
the forming of the satellite bloc in Eastern 
Europe, the fall of China, and the Korean War 
were not predicted. 


P.L. 86-36 
EO 3.3(h)(2) 

Page 263 


"TflS^-Asthe Soviets increased their strategic 
military capabilities, the situation became criti- 
cal. U.S. leaders demanded another Ultra and 
Magic. But th